 
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Journal of the New Zealand Medical Association,
05-May-2006, Vol 119 No 1233
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
How many antibiotic prescriptions are unsubsidised in
New Zealand?
Pauline Norris, Silvia Funke, Gordon Becket, Denise Ecke,
Lillian Reiter, Peter Herbison
Government subsidies for prescriptions are an important
method of ensuring access to medicines people need. In New Zealand (NZ) the
Government pays about 67% of overall pharmaceutical expenditure. While
unsubsidised prescriptions do not contribute to government expenditure,
pharmaceuticals and other medicaments absorb 23% of out of pocket health
expenditure for NZ households.1 This may lead to access problems for some
people; for example those who are just above income and other thresholds. In
addition, the existence of significant numbers of unsubsidised prescriptions
makes any estimations of overall medicine consumption based on PharmHouse
inaccurate. This means it is difficult, for example, to explore the relationship
between antibiotic prescribing levels and resistance. A method of estimating the
number of unsubsidised prescriptions, and correcting for these, is
required.
Unsubsidised prescriptions include prescriptions for drugs
which are never subsidised (e.g. Viagra, Xenical), as well as prescriptions
which cost less than the relevant patient contribution. Until recently, patients
paid a co-payment based on their age (i.e. whether they are under 18 or under 6
years), income (in relation to family size and certain thresholds), recent
health service use (GP visits and/or prescriptions (High Use Health Card [HUHC],
Prescription Subsidy Card)). Whether patients are enrolled in a low cost Primary
Health Organisation now also affects the co-payment they pay. For some drugs
patients pay an additional “manufacturers surcharge”. Prescription
prices are principally determined by the price per day of supply, and the number
of days supplied.2
When medicines are dispensed by community pharmacies they
are recorded in the pharmacy dispensing computer. If the Government is to make
some financial contribution toward the cost of the prescription, an electronic
record of it is also sent to HealthPAC (the organisation which pays government
subsidies to pharmacies). These prescriptions are then included in PharmHouse,
the data-warehousing system which is often used for drug utilisation
research.3–6
When an unsubsidised drug is dispensed, or the price of a
prescription is less than the patient co-payment, the patient pays the full
price, the Government does not subsidise the prescription, and no record of the
prescription leaves the pharmacy. The only repositories of information about
these unsubsidised prescriptions are GPs and community pharmacies. GP databases
cannot tell us whether any given prescription was actually dispensed, and they
do not have pricing data (such as pharmacist fees), so it is difficult or
impossible to determine which prescriptions are subsidised.7
The number of unsubsidised prescriptions is impossible to
determine accurately, since these are not collected by any central agency.
Sutton and Crampton estimated that 20% of prescriptions for adults without
community service cards (i.e. people above the income threshold) were
unsubsidised.8 This was based on the assumption that the proportion of
prescriptions costing less than the subsidy level ($15) was the same for
Community Services Card (CSC) holders (who receive discounted prescriptions upon
presentation of their card) and non-CSC holders. However this may not be the
case because of strategic decisions by doctors, pharmacists, and patients (for
example, patients without CSCs may use stocks of medicines at home, or buy them
over the counter rather than on prescription, if the prescription will not be
subsidised).
While the 20% figure may be appropriate as an overall
estimate, it does not allow us to look at rates of subsidisation for different
drugs. In the study reported here we use data on antibiotic prescriptions from
one town to develop a model which could be used to estimate the number of
unsubsidised prescriptions.
The aim of our study was to determine how many antibiotic
prescriptions were unsubsidised in one town during one year and to develop a
method for estimating the number of unsubsidised prescriptions, using variables
obtainable from PharmHouse data.
MethodsIn this study we collected all antibiotic dispensings
from pharmacy computers of all pharmacies in a small town and compared them with
PharmHouse data for the same pharmacies. We determined the number and proportion
of unsubsidised antibiotic prescriptions, and constructed a model of how to
estimate the number of unsubsidised prescriptions from data contained in the
PharmHouse database.
A town was chosen that was a significant distance from
other towns with pharmacies, and had a small number of pharmacies (but greater
than two to protect commercial information). All pharmacy owners consented to
provide data. We downloaded all dispensings of all prescriptions from each of
the pharmacies in 2002. In some computer systems this was impossible, so data
from a longer period was downloaded. Using Microsoft Access software, data on
dispensings of antibiotics (chemical entities and brand names identified as such
in the New Ethicals Catalogue or the Pharmaceutical Schedule) for the 1
Jan–31 Dec 2002 period was extracted and combined. This produced a list of
all dispensing of antibiotics in the town during the year.
We obtained data from PHARMAC on all subsidised
prescriptions for antibiotics for a 5-year period 1998-2002. Each line of data
represents one or more dispensings and includes the name of the drug, the date
of dispensing, and a number for the pharmacy which dispenses the medicine. Data
for 2002 and for the pharmacies in the town were extracted and summarised using
Microsoft Excel software.
We tabulated the number of dispensings of each drug
from each data source and calculated the proportion of dispensings which were
missing from the PharmHouse data. Using weighted linear regression we also
explored the relationship between this proportion (dependent variable) and
price, length of supply, number of dispensings to CSC patients, to HUHC
patients, to ”J” patients (6–18 year olds), and to
“Y” patients (under 6 year olds) (independent variables).
Regressions were weighted by the number of dispensings from the pharmacy data.
That is, more weight was put on drugs which were more frequently used. It was
crucial that the independent variables were able to be determined within
PharmHouse, so that a feasible and simple method of estimating unsubsidised
dispensings could be developed.
To determine an appropriate price for each drug we
calculated the price of one Defined Daily Dose (DDD) of the most commonly
subsidised formulation of that drug. The most commonly subsidised formulation
was determined from the PharmHouse data. DDDs for all drugs are determined by an
expert committee and listed on the World Health Organisation Collaborating
Centre for Drug Statistics Methodology website (ref: www.whocc.no/atcddd). A DDD is the
quantity of each drug likely to be used daily for its main indication. The price
of one DDD of this formulation was found in the Pharmaceutical Schedule.9
The number of dispensings to CSC patients, to HUHC
patients, to “J”, and to “Y” patients for each drug was
determined in Excel. Length of supply was calculated by dividing the number of
units dispensed by dose per day (e.g. if 12 tablets were dispensed and the daily
dose was 4, then length of supply = 3 days). Many records had missing data about
daily doses. These records were excluded from this calculation. For two drugs
with a small number of dispensings, the daily doses were obviously the result of
a data-entry error, we recorded ‘length of supply’ as missing. For
two drugs administered by injection we recorded length of supply as 1 day.
In the PharmHouse database there was a very small
number of unsubsidised dispensings, which had probably been submitted by
mistake, and another few dispensings where the subsidy from Government was
recorded as less than zero. We ignored both of these in our analysis, since they
are likely to be due to data-entry errors.
The proportion of missing (unsubsidised) dispensings
was regressed on individual variables (price, length of supply, number of
dispensings to CSC patients, to HUHC patients, to ”J” patients
(6–18 year olds), and to “Y” patients (under 6 year olds)).
All the independent variables were then tested together. The least significant
variable was then successively removed from the model.
The data for analysis was one record for each drug,
with the proportion of unsubsidised dispensings and accompanying data. Each of
these was based on a different number of dispensings so contained different
amounts of information and thus could not be treated as equal. So the regression
was weighted by the number of dispensings, which will be proportional to the
inverse of the variance.
Regression as we have used it requires a normal
distribution, however with only 25 data points it is difficult to test for
normality. Simple tests such as a normal probability plot show that deviations
from normality are not too severe, and would be unlikely to affect the result as
regression is robust to most forms on non-normality, and there are no outliers
at the ends of the distribution.
ResultsData from the pharmacy computers in the town included 15,155
dispensings of antibiotics during 2002. Data from PharmHouse for the same period
showed 9768 dispensings of antibiotics. Therefore only 64.4% of dispensings of
antibiotics were subsidised.
The proportion of dispensings which were unsubsidised varied
substantially between drugs (Table 1). For amoxicillin, the most commonly
prescribed antibiotic, 38% of dispensings were unsubsidised. All
benzylpenicillin, and no chloramphenical was subsidised.
Table 1. Unsubsidised prescriptions by drug
When the proportion of missing dispensings was regressed on
individual variables, only price was significant. After successively removing
the least important variable only the estimated price per DDD, number of Y
patients and average length of supply were included in the model and they all
became significant. The equation worked out by this model allows people to put
in values and determine the predicted proportion missing.
The equation is: Data missing
= 0.457 – 0.061*price + 0.000079*Y patients – 0.007*length of
supply
The adjusted R squared value is 0.46, this means that 46% of
the variation in the outcome can be explained by the variables in the model.
This shows how much better the model is at estimating the proportion missing for
each drug, compared to a simple estimate using the average proportion of
unsubsidised prescriptions.
DiscussionOnly 64.4% of antibiotic dispensings in the study town were
subsidised, and therefore captured by the PharmHouse database. This varied
widely by drug. For particular drugs, the proportion of drugs unsubsidised could
be predicted by the price of the drug, the number of days it was prescribed for,
and the number of patients aged under six who received subsidised prescriptions.
A greater proportion of prescriptions are unsubsidised, and
therefore patients are more likely to bear the whole cost, for drugs that are
cheaper, and drugs that are prescribed short-term. This is presumably
appropriate, since other things being equal, these would be less of a burden for
patients. A greater proportion of prescriptions are unsubsidised for drugs where
there are more subsidised prescriptions for children under six. This is likely
to mean that drugs which are prescribed more for young children, are more likely
to be sometimes unsubsidised. This is an interesting finding, which should be
explored in further research.
We have not looked at drugs which are not eligible for
subsidy (i.e. not listed on the Pharmaceutical Schedule). The best way to
estimate the amount of these drugs used would be to approach suppliers.
PharmHouse data cannot provide any information about this.
For our model, ideally estimations of price should be made
for each formulation, rather than using the most commonly dispensed formulation
for each drug. However the data we obtained from pharmacy computers used brand
names, and there were compatability problems between the software programmes
used by the pharmacies, which made identifying formulations difficult.
This study relied on data from only one town, and only one
class of drugs. Further work must be done to validate the model for other
classes of drugs, in other towns. Further work is also needed to investigate the
impact of reduced prescriptions charges in Access Primary Health Organisations
(PHOs) on the level of unsubsidised prescriptions. The proportion of
unsubsidised drugs would be different for other drugs and other towns. For
example, the proportion of unsubsidised drugs is likely to be much lower for
more expensive drugs which are taken for longer periods of time (e.g. statins).
However, the relationship between
price, average length of supply, patient
variables, and the proportion of drugs which are unsubsidised may or may not be
similar for other drugs or other towns, except where PHO funding has changed
subsidy entitlements. Further research is needed into whether this model can
help to estimate the extent of unsubsidised prescriptions.
Previous studies which used PharmHouse data to describe
medicines use are likely to be significant underestimates. We found that only
64.4% of dispensings of antibiotics were subsidised in our study town in one
year. Analysis of PHARMAC data has found there were 2,837,241 subsidised
dispensings of antibiotics in NZ during 2002 (data unpublished). Simply using
the average figure of 64.4% suggests that there may be another 1,568,413
unsubsidised dispensings, thus giving a total of 4,405,654.
Whether previous studies incorrectly estimate regional
variation is debatable. We found that the major drivers of the extent of
unsubsidised dispensings were price per day, length of supply, and number of
patients aged under 6 years who received subsidised prescriptions. Price per day
will not vary by region, length of supply probably does not vary very much3 but
the number of children under 6 years receiving prescriptions may, especially
when small regions are examined.
PharmHouse remains a very valuable source of data on
prescribing trends. Many overseas studies on drug utilisation (especially those
looking at how social characteristics such as age, gender and socioeconomic
status affect medicines) are not able to obtain national records. They may rely
on records from samples of general practitioners,10–12 interviews with
patients13,14 or samples of hospital admissions.15 Those that do use national
dispensing databases often have little patient information available. For
example, in Ireland, Odubanjo et al use eligibility for subsidised medical
services to divide the population into ‘relatively affluent’ or
‘relatively deprived’.16
The addition of National Health Index (NHI) numbers to
PharmHouse data will allow studies of the impact of patient demographics on
prescribing to be carried out at a national level in New Zealand. However, the
exclusion of unsubsidised prescriptions from PharmHouse is currently a major
limitation of this data source. The model we developed in this study may help to
overcome this, however.
Author information:
Pauline Norris, Senior Lecturer, School of Pharmacy; Silvia Funke, PGCert
candidate, School of Pharmacy; Gordon Becket, Senior Lecturer, School of
Pharmacy; Denise Ecke, PGCert candidate, School of Pharmacy; Lillian Reiter,
PGCert candidate, School of Pharmacy; Peter Herbison, Statistician and Associate
Professor, Department of Preventive and Social Medicine; University of Otago,
Dunedin
Acknowledgements: We
thank the pharmacy owners who gave us data; PHARMAC (especially Jason Arnold)
for access to PharmHouse data; Lance Elder and Dick Martin for assisting Lillian
with data-handling; and the School of Pharmacy for providing Postgraduate
Certificate in Pharmacy Scholarships to Silvia, Denise, and Lillian.
Correspondence:
Pauline Norris, School of Pharmacy, University of Otago, PO Box 913, Dunedin.
Fax: (03) 479 7034; email: pauline.norris@stonebow.otago.ac.nz
References:
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| | Current
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads) Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals |
|