The Effects of Mandatory Prescribing of Thiazides for Newly Treated, Uncomplicated Hypertension: Interrupted Time-Series Analysis

Background The purpose of our study was to evaluate the effects of a new reimbursement rule for antihypertensive medication that made thiazides mandatory first-line drugs for newly treated, uncomplicated hypertension. The objective of the new regulation was to reduce drug expenditures. Methods and Findings We conducted an interrupted time-series analysis on prescribing data before and after the new reimbursement rule for antihypertensive medication was put into effect. All patients started on antihypertensive medication in 61 general practices in Norway were included in the analysis. The new rule was put forward by the Ministry of Health and was approved by parliament. Adherence to the rule was monitored only minimally, and there were no penalties for non-adherence. Our primary outcome was the proportion of thiazide prescriptions among all prescriptions made for persons started on antihypertensive medication. Secondary outcomes included the proportion of patients who, within 4 mo, reached recommended blood-pressure goals and the proportion of patients who, within 4 mo, were not started on a second antihypertensive drug. We also compared drug costs before and after the intervention. During the baseline period, 10% of patients started on antihypertensive medication were given a thiazide prescription. This proportion rose steadily during the transition period, after which it remained stable at 25%. For other outcomes, no statistically significant differences were demonstrated. Achievement of treatment goals was slightly higher (56.6% versus 58.4%) after the new rule was introduced, and the prescribing of a second drug was slightly lower (24.0% versus 21.8%). Drug costs were reduced by an estimated Norwegian kroner 4.8 million (€0.58 million, US$0.72 million) in the first year, which is equivalent to Norwegian kroner 1.06 per inhabitant (€0.13, US$0.16). Conclusions Prescribing of thiazides in Norway for uncomplicated hypertension more than doubled after a reimbursement rule requiring the use of thiazides as the first-choice therapy was put into effect. However, the resulting savings on drug expenditures were modest. There were no significant changes in the achievement of treatment goals or in the prescribing of a second antihypertensive drug.

(€190 million), which is NKR 330 per inhabitant (€40). Norway is a country where the use of thiazides is particularly low and the use of expensive alternatives particularly high, and this has been an issue of concern. In 2004, the Norwegian Medicines Agency introduced a regulatory measure that obliged physicians to use thiazides as first-choice drugs for the treatment of uncomplicated hypertension, unless medical reasons suggested otherwise. The effect of this regulatory measure on prescribing habits is uncertain

Methods
We will conduct in interrupted time series analysis of prescribing of antihypertensive drugs to patients that are started on such medication in 60 practices in the Oslo-area in Norway. The primary outcome measure will be the proportion of non-thiazide prescriptions made for persons who are being started on medication for antihypertensive therapy. 2

Background
Antihypertensive medication is one of the major drug-expenditures for the national reimbursement scheme for drugs in Norway, adding up to nearly NKR 1.5 billion per year (€190 million) [1], which is NKR 330 per inhabitant (€40). An estimated 6-10% of the Norwegian population are being treated for hypertension [2].
There are several drugs to choose from for the treatment of hypertension, and in terms of effectiveness there is little or no difference between them [3,4]. One exception is low-doze thiazide-diuretics, which seem to protect better against developing heart failure than most other drug-classes [3,4]. From a public health and economic perspective the major difference between the drugs is undoubtedly the price. Many of the non-thiazide drugs cost tenfold of what thiazides do [2].
Norway is a country where the use of thiazides is particularly low and the use of expensive alternatives particularly high [2], and this has been an issue of concern for the Ministry of Health as well as Norwegian parliamentarians [5]. Substantial savings are expected to result from the new reimbursement rule, however other factors than the anticipated increase in thiazide use may affect this outcome. For instance, some critics believe that thiazides are less effective in controlling blood pressure and that the need for add-on drugs will outweigh the economic gains.
The effects of health policy interventions at national or regional level are often poorly evaluated. Sometimes it is not feasible to conduct rigorous evaluations, such as a randomised controlled trial, due to pressure to implement quickly and widely. Sometimes the nature of the intervention is such that a control group is virtually impossible to establish, as with mass media campaigns, for example.
If a controlled designed evaluation is unfeasible, a before-and-after study may be an alternative. However, if measurements are only made once before and once after the intervention, the findings may be misleading. If, for instance, the prescribing of a drug was measured before and after an intervention and a difference was observed, this could simply reflect an ongoing trend, independent of the intervention. Another explanation for an observed change could be that events coinciding with the intervention had an influence on the prescribing of the drug. One way of increasing the robustness of a before-and-after analysis is to collect several data points before and after the intervention -an Interrupted time series (ITS) analysis.
We want to investigate whether changes in prescribing of thiazides for hypertension have occurred and whether the change can be attributed to the introduction of the new reimbursement rule. Also, we want to explore possible changes in proportions of people with blood pressure levels above recommended targets and the use of add-on drugs.

Design
Interrupted time series analysis, with 11 measurements at monthly intervals before and after the intervention. We will incorporate a transition period from December 2003 through Participants 60 General practices in the Oslo-area, Norway. The sample will be drawn from three different practice populations. 20 practices will be drawn from the practices that were randomised to the intervention group in trial we recently conducted that addressed prescribing of antihypertensive drugs [6]. 20 other practices will be drawn from the control group in the same trial. Finally, 20 practices will we drawn among the practices that did not participate in the trial.

Intervention
A new rule from the Norwegian Medicines Agency, which only allowed for reimbursement of thiazides for patients started on medication for uncomplicated hypertension, in this case 4 defined as not having hypertensive organ damage, gout, reduced glucose tolerance or untreated diabetes. If there were medical reasons not to choose a thiazide other drugs would be reimbursed, on the condition that rational for this was noted in the medical record.
Information about the new rule was circulated to all physicians through the January edition of the bulletin of the Norwegian Medicines Agency [7]. The rule was also widely debated in the media.
No specific sanctions against non-compliant physicians were announced, but it was said that the National Insurance Administration, which manages the reimbursement of drugs, were about to increase their monitoring of adherence to reimbursement rules. There is no reason to believe that the intervention may have affected data-collection.

Data-collection
We will collect data from the practice medical records. The data will include all prescribed antihypertensive drugs from three years before and one year after the reimbursement rule was implemented.
For all patients that have been given a prescription for an antihypertensive drug, we will also collect information on cardiovascular diagnoses (ICPC-codes K74-80, K84, K99) and prescribing of cardiovascular drugs (ATC-code C01). This is to identify patients who may be taking the drugs for other reasons than hypertension. We will also extract blood-pressure measurements, and only patients with a recorded measurement above 140/90 will be included in the main analyses.
Because antihypertensive drugs are also prescribed to treat migraine or hyperthyroidism, we will identify patients with these diagnoses (N89, T85) and prescriptions for drugs to treat these conditions (N02C, H03B). If these diagnoses or prescriptions are recorded the patient will be excluded from the analysis.
The new reimbursement rule only includes patients treated for uncomplicated hypertension, defined in this case as not having target-organ damage, goat, decreased glucose tolerance or untreated diabetes. In order to identify patients with these conditions we will extract information on relevant diagnoses (K87, T92, T89, T90) and prescriptions (M04, A10). We will not attempt to identify patients with decreased glucose tolerance, as we do not believe that the data we extract will enable us to do so in a reliable way. Patients with a recorded glucose measurement ≥ 7 mmol/l and no recorded prescriptions for antidiabetic medication or other diabetes-aids, will be considered as having untreated diabetes, and thus excluded from the main analyses.

Outcome measures
The primary outcome measure will be the proportion of non-thiazide prescriptions made for persons who are being started on medication for antihypertensive therapy. We will conduct a subgroup analysis where all diabetes-patients are excluded (patients with untreated diabetes are excluded from the main analysis). We will also conduct subgroup analyses comparing the three groups of practices that participate in the study.
In addition, the following outcomes will be measured: -The proportion of patients that have not reached recommended blood pressure goals among all that started treatment in the year before vs. the year after the intervention.
-The proportion of patients started on a second antihypertensive drug year before vs.
the year after the intervention.
-The proportion of patients with a diagnosis of heart failure not being prescribed an ACE-inhibitor plus the proportion of patients with diagnosis of coronary heart disesase not being prescribed a beta-blocking agent, year before vs. the year after the intervention. (Proxy measure for inappropriate prescribing).
These are objective outcomes, thus no blinding or reliability testing of outcome evaluation is considered necessary.

Ethics
We will ask for consent from all the participating practices. Physicians and patients will not be identifiable. No ethical approval is needed. We will apply the Norwegian Data Inspectorate for approval for handling of the data.

Sample size and statistics
Based on data from a study we have conducted [6] we estimate that an average general practitioner starts blood pressure lowering therapy on 10 patients per year. In the same study the average number of physicians per practice was 2.7. Thus, we can expect 20-30 new patients per practice per year. In order to detect a 25% relative reduction, with a power 80% and a statistical significance level of 5%, on the prescribing of non-thiazide drugs for hypertension, a sample of 60 practices should suffice, according to expert advice (Craig R. Ramsey, University of Aberdeen).