ORIGINAL ARTICLES AND REVIEWS
Francesca ValentI; Laura DeromaII; Antonella FranzoI, III; Michele GobbatoI; Giorgio SimonIII; Luigi CancianiIV; Loris ZanierI
IDirezione Centrale Salute, Integrazione Sociosanitaria, Politiche Sociali e Famiglia, Regione Autonoma Friuli Venezia Giulia, Udine, Italy
IIAzienda Ospedaliero-Universitaria di Udine, Udine, Italy
IIIAzienda per l'Assistenza Sanitaria n. 5 "Friuli Occidentale", Pordenone, Italy
IVDistretto Sanitario di Codroipo, Codroipo, Udine, Italy
BACKGROUND. Patient referrals to outpatient health services may affect both health outcomes and health expenditures. General practitioners (GP) have a crucial role in driving the use of outpatient services and recognizing factors which affect referrals is important for health managers and planners.
OBJECTIVES. We investigated patient- and physician-related determinants of patient referrals in an Italian region.
METHODS. This was cross-sectional study based on the individual linkage of administrative databases from the health information system of the Friuli Venezia Giulia region. For each GP of the region, the association of the number of patient referrals to different types of outpatient services with the proportion of patients with chronic conditions, with the number of hospital admissions and drug prescriptions in 2012, and with GP's characteristics was investigated through multilevel multivariable Poisson regression models.
RESULTS. Some chronic conditions (e.g., cancer, autoimmune diseases, endocrine diseases, digestive system diseases) were positively associated with the number of referrals, as were hospital admissions and drug prescriptions. Time since GP's graduation was inversely related with referrals.
CONCLUSION. Patient complexity and GP's experience affect referral rates. These factors should be considered in case of a reorganization of the general practice structure in Friuli Venezia Giulia.
Key words: general practitioners, outpatient health services, ambulatory care, chronic diseases
Outpatient service use in a population may depend on a number of factors. Socio-demographic (e.g., sex, insurance status, education) and geographic factors (e.g., area of residence) were shown to affect outpatient service utilization and equity in Iran . In Germany, supply-side factors were shown to account for almost half of the inequalities in access , where demand-side aspects accounted for less than one third and morbidity, in particular, for about 5%. A recent Italian study found some inequity in the use of outpatient specialist care, basic medical tests, and diagnostic services in favor of the well-off , partly explainable because, for accessing those services, even in public hospitals and clinics, there is a small flat rate copayment which may discourage some low income patients from seeking care in cases they judge not strictly necessary.
General practitioners (GPs) play a crucial role in the utilization of outpatient health services. In fact, they may refer patients to specialist care either to obtain one-time advice or to receive continuous feedback on the management of complex cases, especially in situations where their knowledge or confidence are limited. Some substitution effect  cannot be excluded as well: Italian GPs earn a large part of their income through a fixed payment they receive for each patient, regardless of patient's complexity and of the workload associated with the actual patient's treatment and they may be induced to seek a reduction of their workload through the frequent referral of patients to hospital outpatient services.
Referrals to outpatient services may affect not only patients' health outcomes but also health expenditures. Thus, recognizing factors that affect referral patterns, and thus forecasting future demands and identifying sources of inequity, is important to health managers and planner.
In Friuli Venezia Giulia, an Italian Northeastern region with 20 health districts, approximately 1000 GPs and one million inhabitants > 15 years of age, a cohort study had highlighted some heterogeneity in care profiles adopted by GPs among persons with diabetes . However, factors associated with referrals to outpatient services have never been explored. In the administrative databases of the regional health information system, the records of all the outpatient services provided by public hospitals and clinics at the request of a GP are registered. Individual information contained in those databases can be linked through a univocal anonymous key to a variety of additional health-related data. Therefore, we decided to investigate patient- and physician-related determinants of GP's attitudes towards referring patients to outpatient health services in Friuli Venezia Giulia, analyzing the information available from the regional administrative health system databases.
In this cross-sectional study, we used various databases from the regional health information system as sources of information: the GPs database, the outpatient service database, the drug prescription database, the hospital admission database, and the database of exemptions from medical payment. Each person having contacts with the regional health system is identified by a univocal anonymous key, which is the same across all databases and allows deterministic record linkage on an individual basis. We linked on an individual basis all those datasets for the year 2012.
In that year, 980 GPs were active in the Friuli Venezia Giulia region. We attributed patients to each GP based on the actual number of days they were under the care of that physician in 2012. For each patient we used hospital admission, outpatient health service, drug prescription and exemptions from medical charges data to determine whether the person was affected by chronic conditions, according to the criteria described by the Local Health Agency of Brescia, Italy . In short, a person was defined as affected by a certain chronic condition in 2012 if he/she satisfied, in that specific year, at least one of the following conditions: being exempt from medical charges because of that particular chronic condition; being prescribed drugs specifically used for the treatment of that chronic conditions; being discharged from the hospital with a diagnosis related to that chronic condition; undergoing outpatient visits or exams related to that chronic condition.
The following conditions were assessed: chronic lung diseases, chronic heart failure, other cardiovascular diseases (includes hypertension, vascular diseases, and heart diseases), diabetes mellitus, disorders of lipid metabolism, HIV, chronic kidney disease, rare diseases, cancer, digestive system diseases (includes gastroesophageal, enteric, and hepatic diseases), autoimmune diseases (includes rheumatoid arthritis, systemic erythematosus lupus, systemic sclerosis, Sjogren's disease, ankylosing spondylitis, myasthenia gravis, Hashimoto's thyroiditis, autoimmune hemolytic anemia), endocrine diseases (includes acromegaly and gigantism, diabetes insipidus, Addison's disease, hypo- or hyperparathyroidism, acquired or congenital hypothyroidism, Cushing's disease, Basedow's disease, hypophysary dwarfism), neurological diseases (includes epilepsy, Parkinson's disease, Alzheimer's disease, multiple sclerosis, neuromyelitis optica, dementia), transplantation, delivery.
For each GP, we calculated the following variables:
the number and proportion of patients affected by each of the above listed conditions;
the weighted population of patients, where each patient was weighted according to age as follows: 15-24 years 0.390; 25-44 years 0.650; 45-64 years 1.560; 6574 years 2.177; > 75 years 2.074 ;
the number of overall referrals to outpatient health service in 2012;
the number of particular outpatient prescriptions (specialty visits, laboratory services, radiology services, and nuclear medicine diagnostic services) in 2012;
the number of hospital admissions per weighted patient in the population of patients in 2012;
the overall number of drug prescriptions per weighted patient among the patients in 2012.
For each GP, we calculated the overall number of referrals to outpatient health services per 1000 weighted patients as an indicator of the "prescribing attitude" of the GP. Then, we investigated the relation between the number of outpatient health services, both overall and of specific types, and characteristics of the GPs and of their patients through multivariate Poisson regression models in which GPs were clustered in the 20 health districts. Five multivariate models were built: one for all outpatients health service, and one for each of the following outpatient prescriptions: specialty visits, laboratory services, radiology services, and nuclear medicine diagnostic services. The weighed population of each GP was used as the offset variable in the models, with a logarithm link function. The covariates included in each model were the proportions of patients with each chronic condition, the number of hospital admissions, the number of drug prescriptions, the age of the GP and the years since graduation. The association of each outcome with the covariates was expressed through the rate ratios (RR) and 95% confidence intervals (95% CI). P-values < 0.05 were considered statistically significant. All analyses were conducted using SAS Enterprise Guide v4.3 (SAS Institute Inc., Cary, NC, USA).
The distribution of the patients of the 980 GPs active in Friuli Venezia Giulia in 2012 is illustrated in Table 1. On average, GPs had a weighted patients population which was larger than the non-weighted one. Cardiovascular diseases, cancer, diabetes, and disorders of the lipid metabolism were the conditions affecting the largest proportions of patients.
The distribution of referrals to outpatient health services and of the patients with such prescriptions is shown in Table 2. On average, among the GPs of the Region, over 500 persons/1000 weighted patients were prescribed at least one outpatient health service in the year, for an average of more than 11 000 prescriptions/1000 weighted patients, the majority being laboratory exams.
The association between the overall number of referrals to outpatient health services by each GP, the number of patients with chronic conditions, and demographic characteristics of the GPs are presented in Table 3. The number of patients with rare diseases, cancer, diseases of the digestive system, autoimmune diseases, endocrine diseases, of transplanted patients and of women who delivered a baby were positively associated with an increased number of prescription of outpatient health services. Hospital admissions and drug prescriptions per weighted patient were also significantly associated with outpatient health service referrals. On the other hand, the number of patients with chronic heart failure was inversely related with referrals to ambulatory care. Time since the GP's graduation was also inversely related with referrals to ambulatory care.
The magnitude of significant associations with the referrals to specific types of outpatient health services is shown in Table 4.
In this study, consistently with previous research , we found that the distribution of chronic conditions of the patients of a GP influences the GP's attitude towards referring to outpatient heath services. In particular, in our region, the larger the number of patients with cancer, diseases of the digestive system, autoimmune, endocrine, and rare diseases, of transplanted patients and of pregnant women and the more GPs tend to refer patients to hospital outpatient care. The effect of patients' comorbidities on the use of ambulatory services, however, is not uniform across different services. For example, the number of patients with cancer was positively related with laboratory exams, radiology exams, and in particular nuclear medicine exams, which are part of the follow-up of neoplastic patients, but not with visits. This means that, apart from oncologic follow-up visits, which in our region may be provided by the hospital oncology clinics with no need of a GP's prescription, patients with cancer do not increase the referral of patients to other hospital outpatient specialty visits. The number of patients with digestive, endocrine, and autoimmune diseases, on the other hand, are associated with increased prescriptions of visits, laboratory exams, and radiology exams, indicating that these patients may either pose to GPs a quite cumbersome workload, or require a particularly complex management, or be difficult to manage if GPs are not sufficiently trained. Some diseases were associated with only one type of ambulatory service. For example, the number of patients with chronic lung disease resulted associated with increased prescriptions of radiology exams only, or the number of patients with HIV was only associated with increased prescriptions of laboratory exams. This information can be used as a starting point for judging the appropriateness of the present system for paying GPs in Italy. Flat payment on a per-patient basis may not reflect the actual GP's workload. Even weighing patients according to age may not reflect entirely the patients' complexity. Taking into account the distribution of chronic conditions of patients might help estimating the real workload and assigning funding accordingly.
The relevance of the associations that we detected between certain chronic diseases and the prescriptions of outpatient services can be fully appreciated by taking into account not only the magnitude of the associations, but also the frequency of the diseases among the patients of the regional GPs. In fact, despite the fact that rare diseases and transplanted organs are associated with and excess rate of prescriptions of 2.0 and 4.3%, respectively, these conditions are quite rare. Thus, only a limited amount of prescriptions can be attributed to those conditions. On the other hand, autoimmune and endocrine diseases were not only associated with high excess rates of prescriptions (5.2 and 4.3%, respectively), but they were also quite common among the patients of the regional GPs. For the same reason, the impact of cancer patients on prescriptions is relevant, despite a modest RR, because the prevalence of this condition in very high.
The result that the number of patients with some chronic conditions (e.g., diabetes, heart diseases) was not associated with referrals to ambulatory specialty care was unexpected. In fact, for example, GPs should be able to manage the recommended periodic testing of persons with diabetes . Using the regional health information system, in the near future we will need to investigate whether the persons with diabetes residing in Friuli Venezia Giulia undergo the appropriate testing according to the recommended schedules . If the guidelines are largely disattended, the Regional Health Authority will have to promote integrated care pathways which establish the role of GPs in the management of these patients.
Hospital admissions and drug prescriptions were positively associated with ambulatory care use. The relation between these three variables is multifaceted. In fact, a GP who refers many people to hospital outpatient services might be able to manage patients reducing unnecessary hospital admissions. On the other hand, a GP may need to prescribe ambulatory care services more frequently than others if his patients have particularly serious comorbidities and are, consequently, in greater need of hospital care. In addition, part of the prescriptions of outpatient care services might have been suggested by medical specialists who treated patients with severe or complex conditions, with the GP just filling out the prescriptions. Thus, both hospital admissions and the prescriptions of outpatient services and drugs can be interpreted as markers for a highly complex patient population. In our study, however, we adjusted our analyses for the proportions of patients with chronic conditions and so confounding due to comorbidities is at least partly accounted for. Thus, the positive association of outpatient service use, drug prescriptions and hospital admissions could suggest that some GPs have a more evident generic attitude towards "prescribing" than others. However, since we could adjust the analyses for the frequency of the chronic conditions but not for their severity, we cannot exclude that some GPs are frequent prescribers because their patients have more severe diseases.
Differently from Chan and Austin , physician's age was not associated with referral attitude. In our study, however, a characteristics of GPs which seems to influence the referring attitude was years since graduation. GPs who graduated late are "old" if we consider age but "young" if we consider work experience. We found that the more experienced physicians tend to prescribe less outpatient services, regardless of the complexity of the population of patients. This may depend on a greater self-confidence level of these GPs and a smaller need for external aids to diagnosis and management of diseases. Less experienced physicians might be in greater need of training and guidelines. In other settings, GPs reported knowledge deficits and little confidence in the management of chronic conditions [10-14]. In addition to offering GPs high quality training and continuing medical education programs, experimenting new approaches to general practice may improve the quality of care provided by GPs as well as increase efficiency. A recent study showed that Italian patients would prefer a "primary care center" with many diagnostic facilities rather than a "solo GP" model or a "group general practice" . New organizational solutions providing different professional skills and diagnostic services in the same practice and impacting on both the GPs' confidence and waiting times  could be tested. From our study, prescriptions within each health district resulted moderately correlated with each other, suggesting that the district policies influence the GPs' behaviors. If a novel approach to general practice is to be implemented in Friuli Venezia Giulia, the health districts should take an active role in its organization.
The results of this study should be interpreted in the light of a number of limitations. First, we lack data on referrals to private specialty care. Although in the United Kingdom private referral rates were positively associated with the NHS referrals , in our region more than one fourth of the health expenditure is outof-pocket  and we have no information regarding the relation between private and public ambulatory referrals.
Then, we lack information of GPs' actual workload (i.e., the number of visits or contacts with their patients) and on the distribution of socioeconomic variables among a GP's patient population, two factors that may influence referral rates. Thus, our findings may contain residual confounding due to unmeasured variables.
Further analyses may identify additional factors driving patient referral rates in Friuli Venezia Giulia. However, our findings recognized some factors that should be taken into consideration in a possible reorganization of the regional general practice.
1. Doroh VR, Hatam N, Jafari A, Kafashi S, Kavosi Z. Determinants of outpatient services utilization in Shiraz. J Community Med Health Educ 2012;3:216. DOI: 10.4172/2161-0711.1000216.
2. Ozegowski S, Sundmacher L. Understanding the gap between need and utilization in outpatient care-the effect of supply-side determinants on regional inequities. Health Policy 2014;114:54-63. DOI: 10.1016/j.healthpol.2013.08.005.
3. Glorioso V, Subramanian SV. Equity in access to health care services in Italy. Health Serv Res 2014;49:950-70. DOI: 10.1111/1475-6773.12128.
4. Rizzi L, Gobbato M, Tillati S. GLMM models to analyze care profiles adopted on a cohort of diabetics. Politiche sanitarie 2013;14:75-84. DOI: 10.1706/1310.14474
5. ASL Brescia. La presa in carico dei malati cronici nell'ASL di Brescia: monitoraggio BDA 2003-2006. Settembre 2012. Available from: www.aslbrescia.it/media/pdf/pubblicazioni/malati_cronici_2009.pdf.
6. Italia. Ministero della Salute. Riparto disponibilità finanziarie per il Servizio sanitario nazionale nell'anno 2013. Richiesta di intesa alla Conferenza Stato-Regioni. Available from: www.sossanita.it/doc/2013_12_propRIPARTO_FSN2013.pdf.
7. Chan BT, Austin PC. Patient, physician, and community factors affecting referrals to specialists in Ontario, Canada: a population-based, multi-level modelling approach. Med Care 2003;41:500-11.
8. Harris MF, Zwar NA. Care of patients with chronic disease: the challenge for general practice. Med J Aust 2007;187:104-7.
9. American Diabetes Association. Standards of medical care in diabetes-2014. Diabetes Care 2014;37(Suppl. 1):S14-80. DOI: 10.2337/dc14-S014.
10. Thepwongsa I, Kirby C, Paul C, Piterman L. Management of type 2 diabetes: Australian rural and remote general practitioners' knowledge, attitudes, and practices. Rural Remote Health 2014;14:2499.
11. Johnston CL, Maxwell LJ, Maguire GP, Alison JA. How prepared are rural and remote health care practitioners to provide evidence-based management for people with chronic lung disease? Aust J Rural Health 2012;20:200-7. DOI: 10.1111/j.1440-1584.2012.01288.x.
12. Murray S, Lazure P, Schroter S, Leuschner PJ, Posel P, Kellner T, Jenkins RD. International challenges without borders: a descriptive study of family physicians' educational needs in the field of diabetes. BMC Fam Pract 2011:11;12-27. DOI: 10.1186/1471-2296-12-27.
13. Chen Q, Zhang X, Gu J, Wang T, Zhang Y, Zhu S. General practitioners' hypertension knowledge and training needs: a survey in Xuhui district, Shanghai. BMC Fam Pract 2013:28;14-6. DOI: 10.1186/1471-2296-14-16.
14. Tan M, Holloway RH, Lange K, Andrews JM. General practitioners' knowledge of and attitudes to inflammatory bowel disease. Intern Med J 2012;42:801-7. DOI: 10.1111/j.1445-5994.2011.02586.x.
15. Seghieri C, Mengoni A, Nuti S. Applying discrete choice modelling in a priority setting: an investigation of public preferences for primary care models. Eur J Health Econ 2014;15:773-85. DOI: 10.1007/s10198-013-0542-8.
16. Mulvaney C, Coupland C, Wilson A, Hammersley V, Dyas J, Carlisle R. Does increased use of private health care reduce the demand for NHS care? A prospective survey of general practice referrals. J Public Health (Oxf) 2005;27:182-8.
17. Istituto Nazionale di Statistica. NoiItalia. 100 statistiche per capire il Paese in cui viviamo. Roma: ISTAT; 2014. Spesa sanitaria delle famiglie. Available from: http://noiitalia.istat.it/fileadmin/user_upload/allegati/34.pdf.
▲ Address for correspondence:
Regione Autonoma Friuli Venezia Giulia, Direzione Centrale Salute, Integrazione Sociosanitaria, Politiche Sociali e Famiglia, Servizio Epidemiologico
Via Pozzuolo 330
33100 Udine, Italy.
Received on 20 March 2015.
Accepted on 13 May 2015.
Conflict of interest statement: None.
Authors' contributions: Francesca Valent participated in the study design and in analysis of the data and wrote the manuscript; Laura Deroma participated in the study design and in analysis of the data and critically revised the manuscript; Antonella Franzo and Michele Gobbato contributed to the analysis of the data and critically revised the manuscript; Giorgio Simon and Luigi Canciani provided inputs to the analysis and critically revised the manuscript, Loris Zanier participated in the study design and critically revised the manuscript.