The Portuguese NHS is a publicly funded single-payer system, with each citizen being registered in a primary care practice and having an assigned family physician.15 Some of these family physicians participate as volunteers in the Portuguese Sentinel Practice Network,16 each of whom provides a cohort of their registered patients. This allows the network to maintain an open cohort of patients that is reasonably stable in each year and means the incidence of health problems can be calculated in this sample of the Portuguese population. The network was set up to conduct weekly surveillance of communicable and non-communicable diseases, and has also been used for observational epidemiological research to answer specific questions (satellite studies).

Between January 2014 and December 2015, the network expanded the information being reported on the incidence of hypertension for surveillance purposes, to conduct a continuous notification cross-sectional study among new cases of hypertension notified in the cohort.

In 2014 and 2015 the population under observation consisted of 35535 individuals, distributed between 82 family physicians participating in the Portuguese Sentinel Practice Network. Participating physicians were asked to notify all incident cases of hypertension. Hypertension was defined using the criteria adopted by the Portuguese Directorate-General of Health (DGS) (systolic blood pressure of 140 mmHg or higher or diastolic blood pressure of 90 mmHg or higher on several separate occasions).17

Data were collected using paper or online forms. Notifications included information on patient age at diagnosis, gender, pharmacological treatment (entered in a free text field on the paper form, which was subsequently coded by the investigators as the international nonproprietary name; the online form had a list of antihypertensive drugs available on the Portuguese market by international nonproprietary name, of which up to three could be entered), other treatment measures (free text field), who had issued the initial prescription (the family physician participating in the sentinel practice network or a different physician, who was then specified in a separate free text field; these were subsequently divided in two categories: family physicians and other specialists), and, for prescriptions initiated by other physicians, whether the notifying physician had changed the treatment. The Sentinel Practice Network coordinating team followed up any submissions with incomplete or unintelligible information, contacting the notifying physician to gather missing data.

The main outcome in this study was the proportion of patients who began treatment with each class of antihypertensive drug. Classes were defined using the Anatomical Therapeutic Chemical classification for antihypertensives (C02), diuretics (C03), beta-blocking agents (C07), CCBs (C08), and agents acting on the renin-angiotensin system (C09).18

Secondary outcomes were the proportion of patients for whom initial diagnosis and prescription of therapy were by family physicians (those participating in the Sentinel Practice Network or other family medicine specialists) or by other specialists (hospital and private based); the proportion of patients for whom Sentinel Practice Network physicians altered prescriptions initiated by others; and the proportion of each class of antihypertensive drugs prescribed by family physicians as opposed to other specialists.

Drugs were compared regarding use as single treatment combinations in a given patient and total use (as part of any combinations of antihypertensive therapy – alone, in fixed-dose associations or combined with the administration of other drugs as separate pills).

To estimate the proportion of patients beginning treatment with each class of antihypertensive drug with 5% precision and 95% confidence, assuming as a worst-case scenario that 50% of patients would begin with a given class, a minimum sample size of 384 cases of hypertension was calculated. Given the notification rates in previous years of 206 new cases of hypertension per year, we estimated that two years of continuous notifications would be needed to achieve our target sample size.

Proportions of each antihypertensive class prescribed as initial therapy were estimated with the respective 95% confidence interval (CI). Patient distribution regarding gender among family physicians and other specialists was compared using Fisher's exact test and age distribution was compared using the t test. Prescription patterns of family physicians and other specialists were compared using multivariate logistic regression analysis adjusting for patient gender and age. A level of significance (α) of 0.01 was used as the threshold for statistical significance to account for multiple comparisons.

The study was approved by the Ethics Committee of the Portuguese National Health Institute Doutor Ricardo Jorge.

The clear majority of patients with newly diagnosed hypertension were prescribed pharmacological treatment. Two-thirds of patients were started on a single drug, with physicians preferring, in decreasing order, ACE inhibitors, thiazide-like diuretics and ARBs. Most patients who were prescribed more than one drug were started on fixed associations, of which ACE inhibitors with thiazide diuretics or with CCBs were preferred. Loop and potassium-sparing diuretics were rarely prescribed in first-line treatments, and there were no reports of the use of direct renin inhibitors.

Most diagnoses were made by the patient's family physician. Family physicians did not differ significantly from other specialists regarding the use of most classes of antihypertensives, but used less beta-blockers and loop diuretics. When other specialists initiated therapy, it was usually continued by family physicians.

By prospectively collecting data within the Sentinel Practice Network cohort, we were able to gather accurate data regarding drugs being prescribed for new cases of hypertension, avoiding recall bias and administrative inaccuracies. In most cases, the physician reporting the case was the one who had made the diagnosis and therapeutic decision a few moments before. For cases in which the diagnosis had been made by other physicians, information was gathered from patients when they visited their family physicians to request refill prescriptions or for other medical problems. If needed, additional information could be gathered from the patient's national record, which allows the family physician to view details of prescriptions issued elsewhere, including by hospital and private physicians (electronic prescription has been mandatory since August 2011 to be eligible for reimbursement by the Portuguese National Health Service).19 Having this information available to Sentinel physicians greatly reduced patient recall bias in our study.

Some patients may have been diagnosed with hypertension by other physicians and not visited their family physician during the study period, and were therefore not reported to the Sentinel Practice Network. Hence, the incidence of hypertension may be underestimated and the proportion of diagnoses made by family physicians may be overestimated in our study. However, hypertension incidence was similar to that reported previously in the Sentinel Practice Network,20 and by authors in other settings.21 This suggests that under-reporting was low.

The study did not collect data on disease severity, the presence of comorbidities or contraindications to specific drugs. These could influence the choice of initial treatment, particularly under Portuguese guidelines. However, gathering such information would have increased the amount of time physicians had to dedicate to each notification and the space needed on the paper notification forms. Both would make the study unacceptable to the Sentinel Practice Network. We cannot exclude the possibility that patients with more severe clinical conditions on initial presentation, such as acute cardiovascular events or heart failure, were more often seen and managed by specialists than by family physicians. This may explain the higher prescription of beta-blockers and loop diuretics.

Yet these differences between family physicians and other specialists must be taken as hypothesis-generating only. The comparison of prescribing patterns was a secondary outcome in our study and multiple comparisons were made. Despite having used a statistical threshold of 0.01 instead of 0.05 for significance to account for multiple testing, it is still possible that these differences represent false positive results. We did not define a formal adjustment method for multiple comparisons in the study's methods since we could not have known how many combinations of antihypertensive treatments would be present in our sample. There may also have been other differences, but our study was underpowered to find them.

Although not statistically significant, there were large absolute differences between family physicians and other specialists in the proportion of men diagnosed, mean age at diagnosis, and the proportion of patients treated with a single drug. Our study may have been underpowered to find these differences, as there were relatively few cases diagnosed by other specialists.

Another limitation is that physicians who volunteer to participate in the Sentinel Practice Network may be different from other family physicians, resulting in different prescribing patterns. However, other authors have found these differences to be small.22,23 Also, given that the observed prescribing patterns for Sentinel physicians and other specialists were similar, it is unlikely that they differ significantly from those of other family physicians.

Previous research on prevalent use of antihypertensives had reported greater use of multiple agents, of ARBs relative to ACE inhibitors, and of diuretics, CCBs and beta-blockers.8,24 Greater use of combination therapy than in our study is to be expected, since an analysis of prevalent use will include patients with longer disease duration. More use of diuretics, CCBs and beta-blockers may be explained by the use of these agents mainly as add-on therapy when blood pressure goals are not achieved. Our study showed that ACE inhibitors were used almost two and a half times more than ARBs for initial treatment. This is different from what was observed in previous studies on prevalent use and administrative data on drug dispensing, in which ACE inhibitors were used less often than ARBs.2,8,24 Since ACE inhibitors and ARBs are seldom used in combination in the same patient, comparing prevalent use can approximate the relative proportion of patients treated with each of them. Lower prevalent use of ACE inhibitors may indicate that, after initiating treatment, a significant proportion of patients substitute them for other classes like ARBs. It is also possible that prescription patterns have changed since the previous studies and physicians are now preferring ACE inhibitors over ARBs.

Participants in the Sentinel Practice Network usually did not alter prescriptions issued by other physicians. This may be because they agree with the prescription, as our results show no major differences in prescribing patterns between family physicians and other specialists. It is also possible that family physicians feel they do not have enough information to change prescriptions initiated by specialists, the patient will continue under the responsibility of the other physician, or the patient would be resistant to change.25

Pharmacological treatment employed a large variety of drugs of different classes, and both family physicians and specialists appear to be following Portuguese and European hypertension guidelines for treating new patients.6,7 Since use of ARBs is much lower in incident than in prevalent patients, future studies should focus on how drugs are used after the initial diagnosis, particularly if ARBs are introduced because of adverse reactions, as substitutes or as add-on treatment due to failure of blood pressure control with initial therapy or for other reasons. If Portugal is to reduce use of ARBs, guidelines should include further recommendations on how to manage patients who are not adequately controlled or who experience adverse reactions with the drugs initially prescribed.

Family physicians and specialists had similar prescribing patterns, but the former were the main initiators of antihypertensive treatment. Therefore, efforts to improve quality of care in treating hypertension should be mainly directed at family physicians.