How are reasons for encounter associated with influenza-like illness and acute respiratory infection diagnoses and interventions? A cohort study in eight Italian general practice populations

In this study, the content of family practice is measured with the ICPC [15‐17].

This classifies patient data and clinical activity in the domains of family practice and primary care. It allows classification of the patient’s RfE, the problems and diagnoses managed, the interventions, and the ordering of these data in an ‘Episode of Care’ (EoC) structure [15‐17]. The ICPC has a biaxial structure and consists of 17 ‘chapters’, each divided into 7 ‘components’ (Additional file 1) [17]. The RfE is defined as an agreed statement of the reason(s) why a person enters the health care system and represents the demand for care by that person [16, 17]. An EoC is defined as a health problem from its first presentation by the patient to the family physician, until the completion of the last encounter for it. It encompasses all contact elements related to that health problem [17].

Italian family physicians who belonged to ‘ICPC Club Italia’, an organisation with 12 members that works on the introduction and development of the ICPC in Italy, were invited to take part in the study. During a 4-month period (December 2013 to March 2014), they collected data on patients that they diagnosed as either having an influenza-like illness (ICPC code R80) or a different acute respiratory infection (comprising ICPC codes H71 otitis media, R74 acute upper respiratory infection including rhinitis, rhino pharyngitis, pharyngitis, R75 sinusitis, R76 acute tonsillitis R77 acute laryngitis, R78 acute bronchitis and R81 pneumonia) [13, 14, 25, 26]. We used the ARI and ILI categories as they have been found to be a valid tool for monitoring frequently occurring respiratory diseases [5, 6, 14, 15, 26‐28]. Participating FPs were asked to complete an electronic form (Additional file 2) in an EoC structure based on the ICPC classification. For each EoC, the form prompted FPs to give data on patients’ age and sex, RfEs, the number of encounters for that EoC, procedures, the method of each encounter (at the FP’s practice, by telephone, at the patient’s home), whether the diagnosis was an ILI or an ARI, and whether or not the patient had had a pre-season influenza vaccination. To assist them with their coding, participating FPs were issued with an Italian-language version of the abbreviated, two-page version of ICPC-2 [29]. Completed data collection forms were sent to two independent coordinator centres by email.

As this was neither an interventional nor an observational study on pharmacological treatment, in accordance with local regulations the approval of the ethical committee was not required.

For continuous variables, mean values were calculated, and for categorical variables, percentages were calculated. Adjusted odds ratios (OR) and 95% confidence intervals (CI) were calculated using logistic regression analysis to compare the likelihood of specific symptoms presenting in patients diagnosed with ILI with those diagnosed as having ARI, to compare the symptoms of patients who were given prescriptions related to the respiratory system, were given sick notes, and who requested a return visit. Statistical analysis was performed using Epi-Info v7.1.4 (Center for Disease Control and Prevention, Atlanta, USA).

This study is the first investigation of the distributions of RfEs for ILI and ARI diagnoses made in an Italian primary care setting, collecting data on elements of doctor-patient encounters in an EoC structure. The commonest RfEs recorded for patients diagnosed with ILI were fever, cough and pain. Patients diagnosed with ARI presented most commonly with cough, throat symptoms and fever. Fever and pain were more likely in patients diagnosed with ILI than ARI, while throat symptoms were more likely in patients diagnosed with ARI. Less than half of all patients received a prescription, and fewer than 1% of patients were referred to specialists and/or hospitals or had tests requested. Subjects who had been vaccinated for influenza, and those aged over 50, were less likely to be diagnosed as having an influenza-like illness.

For patients who were given prescriptions, the symptoms tended to be different for ILI and ARI, with headache and generalised pain being commoner in patients who were subsequently diagnosed as having ILI, and cough and throat symptoms being more often seen in patients with ARI. For ILI fever and cough, and for ARI cough, fever and throat symptoms, were the main predictors of return visits for those patients.

Although the ILI patients in this study were not tested for presence of the influenza virus, the weekly ILI incidence recorded by this group is comparable to that of the Italian influenza national sentinel surveillance data (Influnet Italy) [30] for the same time period (Fig. 2). There were no comparable ARI Italian national surveillance data.

Our results are compatible with those of a multinational study which reported that, in individuals with ILI, the two best predictors of a laboratory-confirmed diagnosis of influenza were cough and fever. The authors included eight double-blind, placebo-controlled studies involving 231 study centres in North America, Europe, and the Southern Hemisphere. Of 3,744 subjects enrolled, 2,470 (66%, mean age, 35 years) were laboratory-confirmed to have influenza. Of those, 49.5% were females [31]. In our study, the mean age of ILI subjects was 44 years, and 48.1% were females. Despite a different methodology and study population, both studies confirm that, although FPs are often informally aware of the arrival of influenza virus in the community, their knowledge could be increased with the help of better surveillance and rapid confirmation of infection, especially at the start of an epidemic, when information is scanty. Since it is not feasible for FPs to collect diagnostic specimens from their patients during pandemic influenza [3, 4], the combination of better surveillance with the symptoms of cough and fever could improve the accuracy of FPs in making a clinical diagnosis of influenza.

In a cohort study that took place during the 1998/9 Italian winter epidemic period, 202 FDs performed almost 200,000 visits to 276,000 patients. A total of 6,057 cases of ILI were studied [9]. In contrast to our cohort, the most prevalent systemic symptoms were headache (70.2%) and myalgia/arthralgia (70%), followed by anorexia (59%) and feverishness (35.4%). The most prevalent respiratory symptoms were cough (82%) and sore throat (62.8%). Compared with our data, a much higher proportion of that group of patients received at least one prescription (97.3%), while a similar proportion had received pre-season vaccinations for influenza (5.9%). A higher proportion (4.2%) of that group needed a diagnostic test, specialist assessment or hospitalisation. Many more of the 1989/9 cohort were seen at a home visit (65.7%), but no telephone consultations were recorded.

Another study has shown that linking antibiotic prescriptions to specific diagnoses using the International Classification of Disease and Related Health Problems – Tenth Revision (ICD-10) could reduce those prescriptions [32]. Since FPs are responsible for most antibiotics prescribed to humans, a more specific ability to diagnose different respiratory infections, using both defined FM tools to collect data and appropriate guidelines, may help them better manage these drugs.

In a prospective observational study involving 2,191 ILI and ARI patients (49.8% females) that took place during the 2003/4 Italian winter influenza epidemic period, 508 cases of ILI and 1,683 ARI were gathered [14]. Compared to our population, in that study the percentage of ILI-ARI subjects was higher in the age ranges 5–14 and 45–65, and lower in patients over 65. While one may have expected fewer home visits in that study due to its younger population, 30% of those patients, higher than the 17% in our study.

A cross-sectional study on the ability of 60 primary care physicians to diagnose respiratory diseases found that, out of 235 patients (65.5% females) diagnosed as having ARI [33], the most prevalent respiratory symptoms were cough (90%), followed by fever (50%) and dyspnoea (25%). In that study, FPs were invited to fill out a symptom-based standardised respiratory questionnaire. Their results indicate the highest agreement between the diagnoses of the FPs and the respiratory physicians in ARI (k. 0.53,95% CI 0.46–0.60). Despite a different population and methodology, our study reached a similar conclusion in managing these diseases, with fewer or no referrals to specialists.

In a year-long retrospective study of 439 patients (71% females) seen in primary care because of ARI, the most common symptoms were found to be cough (present in 64% of ARI patients seen), sore throat (55%) and nasal symptoms (47%). FPs ordered rapid testing for group A streptococci in 18% of patients and chest x-rays in 8% of them. Twenty patients were referred to specialist. Clinicians prescribed antibiotics in 213 (49%) of them [34]. The authors concluded that interventions like accurate, reliable pre-visit triage and management, and internet-based medical visits, or E-Visits, which reduce ARI visits, have the potential to decrease inappropriate antibiotic prescribing, reduce the burden of ARI office visits on the health care system and offer more convenience for patients [34, 35]. In our cohort, the percentage of all respiratory prescriptions for 848 ARI patients was 38.8%. FPs referred four patients to specialists/hospital and 632 of them (74.5%), had a clinical examination. It may be that the recording of the patient’s reason for encounter, as well as the doctor’s diagnosis, may have triggered a more appropriate management in our patients. The ICPC is a classification which allows precise ordering of the data elements and concepts within a domain, with unique codes for unique and defined concepts [15, 17].

In a recent prospective study in Denmark, 2,323 ARI patients were diagnosed with either acute pharyngotonsillitis, acute otitis media, acute rhinosinusitis, acute bronchitis, pneumonia or acute exacerbation of COPD, according to the second edition of International Classification of Primary Care (ICPC-2). Less than half of all patients diagnosed with ARI received a prescription, which is lower than the antibiotic prescribing rate for a variety of ARIs in a recent study in Denmark [32]. Their conclusions ‘to improve antibiotic prescribing in general practice, it is important to focus on both the diagnostic process and the prescribing patterns’, are similar to ours.

In our study, some of the patients’ reason for encounters were independent predictors of issuing a prescription in both diseases. This is in accordance with other studies that showed a strong association between patients’ RfEs and the interventions made by their FPs [16, 19, 22, 24].

The study used an internationally validated tool to collect data in general practice [15, 17, 22, 36, 37] on aspects of doctor-patient encounters in an EoC structure, over a complete winter influenza epidemic period. The age and sex profile of the cohort was similar to that of the Italian population as a whole (Additional file 4). The data were collected from patients’ electronic medical records, and many studies suggest that these data do not differ significantly from survey data based on self-report [16, 20, 38, 39].

Of the 1,536 patients whose data were used in the study, 36 were diagnosed as having both ILI and ARI. However, this is unlikely to have affected the analysis, as in each of those patients their ILI and ARI episodes of care were over different time-periods.

We had fewer participating doctors than in some other studies of ILI in primary care [9, 33], and we used a convenience sample of FPs, so our findings may not be generalisable to other Italian FPs. None of the patients diagnosed with ILI had laboratory testing for the influenza virus so some may not have been suffering from influenza. However, this reflects normal clinical practice and is compatible with EISN standards [1, 2], and the weekly incidence profile of ILI in our group was similar to that of national Italian influenza diagnoses over the same time period. Because FPs did not code the EoCs of all the patients that consulted them during the data-collection period, an estimation of the predictive value of symptoms for ILI and ARI diagnoses was not possible.

In Italy 67% of FPs work in an urban area [40, 41], and in our study half of the FPs were rural. At the time of the study, females made up 51.5% of the Italian female population [42], closely comparable with the 51.8% of the population of the participating practices in our study who were female (Additional file 4). In Italy, all citizens are registered with a primary care doctor, so the practice populations represent a cross-section of their local population. The ages of the participating practice populations closely aligned to those of the Italian population, with exception of patients were aged below 20 who were under-represented. This is likely to be because, in Italy, patients aged under 14 years are mainly seen by paediatricians.

The ICPC coding was done by physicians during routine clinical practice, so there may have been some omissions in RfEs, diagnoses and process codes due to individual errors. In addition, FPs may have seen, but failed to code, some ILI or ARI patients. The study was carried out in four out of the 20 Italian regions, with eight FPs. Their patients may be not representative of all the Italian population, so the findings may not be generalisable to that population. Though the study gathered information on ICPC Component 3 (medication, treatment and procedures), participating doctors were not asked to record the kind of medication they prescribed.

Patients who have ILI and ARI can be managed with very low levels of diagnostic testing and specialist referral. The lower levels of ILI in patients aged over 50 may be due to the higher rate of pre-season influenza vaccination in that group. Primary care clinicians in other geographical areas will be able to compare these data with their own activity by using the ICPC coding system.

Knowledge of how the predictors for ILI and ARI compare will help doctors to implement early infection-control strategies and to assess the appropriateness of drug therapy.