Family Practice Advance Access published online on September 28, 2007
Family Practice, doi:10.1093/fampra/cmm059
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Patient psychosocial factors and primary care consultation: a cohort study
a Centro de Salud "El Palo", Distrito Sanitario Málaga, Unidad Docente de Medicina Familiar y Comunitaria de Málaga, Grupo SAMSERAP y redIAPP and Departamento de Medicina Preventiva, Universidad de Málaga, Campus Universitario de Teatinos s/n, 29071 Málaga
b Escuela Andaluza de Salud Pública, Campus Universitario de Cartuja s/n, 18080 Granada
c Departamento de Bioestadística, Universidad de Granada, Avenida de Madrid 11, 18071 Granada
d Departamento de Medicina Preventiva, Universidad de Granada, Avenida de Madrid 11, 18071 Granada, Spain
Correspondence to Juan Angel Bellón, Facultad de Medicina, Departamento de Medicina Preventiva y Salud Pública, Campus Universitario de Teatinos s/n, 29071 Málaga, Spain; Email: JABELLON{at}terra.es
Received 1 December 2006; Revised 22 June 2007; Accepted 22 August 2007.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Background. The combined influence of psychological distress, family dysfunction and social support on primary care consultation (PCC) remains unclear.
Objective. To build an explanatory model of PCC concerning users’ psychosocial factors.
Methods. We undertook a multicentre, prospective cohort study of a random sample of 1141 persons assigned to 113 GPs, belonging to 11 urban health centres in four Spanish cities (Seville, Malaga, Jaen and Granada), of whom 955 (84%) were interviewed in their homes. They were followed up for 1 year and then contacted again. After the second interview, 70 (7.3%) patients were excluded; accordingly, we measured the number of PCC of 885 valid patients using their medical charts.
Results. A multilevel analysis was developed. The null model with three levels showed that 93.29% of the variability was explained by the patients, 1.56% by the GPs and 5.15% by the health centres. We selected a two-level model (patients and health centres) with random effects. The variables used in the multilevel analysis explained 48% of PCC, 36% at the patient level and 12% at the health centre level. Poor mental health (GHQ-28, partial correlation coefficient = 0.28) and family dysfunction (Family APGAR index, partial correlation coefficient = 0.26) were the most predictive variables, whereas social support (Duke-UNC-11, partial correlation coefficient = –0.14) lost significance in the multivariate analysis. Chronic illness seemed less relevant in our study, and only two predisposing factors were included in the equation: age and satisfaction with their doctor.
Conclusions. Mental health and family function were the most important psychosocial factors predicting PCC. More comprehensive identification of psychosocial factors may enhance our understanding of PCC.
Keywords. Family function, mental health, primary health care, social support, utilization of services.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
A systematic review indicated a variety of characteristics associated with frequent attendance.1 The review showed that frequent consulters are highly heterogeneous and have high rates of physical disease, psychiatric illness and social difficulties. The association between mental health and frequent attendance has been documented for most psychiatric disorders,2–5 especially in patients with mood disorders.6 However, doubts arise concerning this association because only a few studies have had a prospective design.7–9
Few studies have explored family function as predictor of primary care consultation (PCC). In one study, less family expressiveness and cohesion and greater family irritability and negation were associated with a rise in PCC.10 In another study,11 higher perceived family criticism predicted more psychosocial, somatic and biomedical visits. Additionally, social support is also associated with the utilization of health services.12 Social support, family function and psychological distress are factors that interact and overlap. One study showed that low social support interacted with psychological distress to increase the use of services,13 and other results supported the primacy of family functioning factors in understanding the associations among social relationships, mental health and health behaviours.14 One of the few studies to explore these factors together15 found that mental health and family dysfunction were variables strongly associated with frequent attendance, whereas the importance of social support disappeared in multivariate analysis. However, this study was cross-sectional in design, making it difficult to draw any conclusions about the direction of any observed association.
A recent systematic review16 indicated that frequent attender studies suffered from methodological problems. Besides the cross-sectional designs, the use of PCC as a categorical variable and the different definitions of frequent attenders hamper comparison of their precision, validity and generalizability. Another methodological concern is the need for multilevel analysis to control the variability due to the individual patient and the different patient groups.17 Unfortunately, the analytical approaches used in previous studies did not properly account for these two sources of variability in PCC estimates.
In our study, we investigated prospectively the association between comprehensive patient psychosocial factors and PCC, defined as a continuous variable, and using multilevel analysis.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Setting
The source population comprised of 1 647 000 patients attending 76 urban health centres in four cities in southern Spain: Seville (32 health centres, 719 000 inhabitants), Malaga (22 health centres, 550 000 inhabitants), Jaen (7 health centres, 113 000 inhabitants) and Granada (15 health centres, 265 000 inhabitants). Each health centre, which covers a population of 15 000–28 000 inhabitants from a geographically defined area, is staffed by GPs, who see patients over the age of 14 years, and primary care paediatricians. The physicians in each health centre work as a group, with extensive primary care teams. The Spanish National Health Service provides free medical cover to 100% of the population. Patients do not pay directly for this service and hence they have no financial constraints on consultation.
Subjects
The eligible population comprised of all subjects aged 14 years or over who had a record at any of the health centres. We chose a random sample of 11 health centres (Seville 4, Granada 3, Malaga 2 and Jaen 2) from among those that, in each city and at the time of the study (1999), had been open for at least 2 years (65 of 76 health centres). In each health centre selected, we chose a random sample of patients (about 100 patients in each health centre) from the lists of all the GPs belonging to each health centre: 416 patients in Seville (41 GPs), 310 in Granada (29 GPs), 210 in Malaga (20 GPs) and 205 in Jaen (23 GPs). The selection of the health centres and the patients was done by computer-generated randomization performed by a person independent of the research team. We selected 1141 individuals, assigned to 113 GPs. Of the 1141 patients, we interviewed 955 patients (84%), who all gave informed consent. Of the 186 patients who were not interviewed, 68 (6%) had moved out of their health centre district, 39 (3.4%) could not be contacted after three visits to their home address, 24 (2.1%) declined to take part in the interview, 20 (1.7%) had incorrect telephone or address details, 15 (1.3%) were disabled (dementia, deafness and learning disability), 12 (1.0%) had died and 8 (0.7%) were not included for other reasons. There were no significant differences in the distribution of sex or age between the patients who were unavailable for interview and those who were interviewed.
Dropouts and final sample
We undertook a multicentre, prospective, cohort study of 955 patients interviewed in their homes between January 1 and June 30, 1999. The patients were followed for 1 year and then contacted again. At this point, we excluded 70 (7.3%) patients: 31 had changed their GP during the follow-up; 28 could not be located after three calls or three visits to their home on different days, weeks and hours; 9 were away for 3 months or more during the follow-up and 2 patients had died. The final sample comprised of 885 patients. Using the method of Snijders18 and from the sample data, the study had a power of 80.9% (with an alpha error of 5%) to detect a correlation of 0.15 between the logarithm of PCC and the square root of the GHQ-28 score. The only significant difference in the distribution of the study variables between the patients who had completed the follow-up period and those who had not concerned their city of residence is as follows: in Seville, 291 (88.7%) patients completed the follow-up, in Granada 258 (96.6%), in Malaga 169 (93.9%) and in Jaen 167 (92.8%); chi-square = 14.094, d.f. = 3, P = 0.003.
Variables
The dependent variable was ‘PCC’. Both patient- and doctor-initiated consultations were included, and the variable was scored as all visits (face-to-face or home visits, during regular hours or out of hours, and for clinical or administrative reasons) to the GP during the 1-year follow-up from the initial interview date, as determined from the patient’s clinical record. Visits to the hospital emergency department and telephone contacts were not included because reliable data were unavailable. We excluded routine antenatal contacts, which overestimate health care problems for women.
All independent variables were recorded by trained interviewers during the course of individual interviews with the patient at his or her home. The independent variables were age, sex, marital status, level of education and employment status according to the classification of the ‘Spanish Institute of Statistics’19; socio-economic level according to an adaptation to social class of the ‘British Registrar General’s Social Classes’ from the ‘National Classification of Occupations’20; number of persons per household; self-reported health (Likert scale, range 1–5); satisfaction with their GP (Likert scale, range 1–5) and travelling time to health centre (1 = less than 5 minutes, 2 = 5–10 minutes, 3 = more than 10 minutes and 4 = homebound); the categories of these variables are shown in Table 1. Data were also recorded on the number and type of reported chronic illnesses (list of 33 chronic health problems cited in a previous study15) according to the pertinent items in the ‘National Health Survey’.21 Moreover, we asked if the patient had been to a healer, to an alternative medicine doctor or to a private doctor during the previous year (response: yes/no). The psychosocial variables measured were mental health, Goldberg's GHQ-28 questionnaire22; family function, Family APGAR23,24 and social support, Duke-UNC-11 scale.25,26 All independent variables were measured at the beginning of the study. At the end of the follow-up, we contacted the patients again to achieve information regarding exclusion criteria.
|
Statistical analysis
Quantitative variables were analysed to rule out skewness and were transformed with Tukey's27 criteria if necessary. Bivariate analysis included Student's t-test, analysis of variance-1 and Pearson's correlation coefficients. All confidence intervals (CIs) were set at the 95% level, and we considered the level of significance as P < 0.05. The logarithm (x + 1) of the number of visits to the GP during the follow-up period was the dependent variable of a multiple linear regression from a multilevel analysis that discerned three levels: patient, doctor and health centre. Patients could be nested in groups (doctor, health centres, districts, etc.) and patients could be similar in the same group; traditional regression models therefore could not be used because the hypothesis of independent observations was not satisfied. A multilevel analysis with a hierarchical model was used to take into account the distribution of the data at different levels to estimate two types of variability, one due to individuals in the study and another due to the groups in which the patients were nested. The independent variables were included in the model using an entrance value of P < 0.15 and an exit value of P > 0.20. These criteria guaranteed that the information lost as a result of exclusion of a variable from the equation was reasonable.28 However, regardless of the P-value, no variable was excluded from the model if the variable modified the coefficients by more than 10%. The process of elimination of chronic illness variables in the multivariate model was as follows: first, we removed the chronic illness variables that were not significant (P > 0.20) and then, we added to the final model, one by one, all chronic illness variables that were not introduced in the model in the first step, in case a variable should be considered. The usefulness of including first-degree interactions in the equation was also considered. All bivariate analyses were run with SPSS 11.5 and we used MLwN 1.1 for the multilevel analysis.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
The mean age of the 885 patients was 47.8 years (95% CI, 46.6–49.0; SD = 18.8), of whom 508 (57%) were female, 68% were married, 57% had a primary school certificate or lower and 68% were from social classes IV or V. The mean number of consultations during the follow-up was 5.0 (95% CI, 4.6–5.4; SD = 6.1). Further details are given in Table 1.
Higher satisfaction, poor perceived health, retired status and a lower level of education were significantly related with higher rates of use. Lower social class was also related with higher rates of use, although not significantly. Widows and widowers and separated persons used primary care services more frequently than did unmarried and married persons. Utilization was higher at health centres in Malaga than in the other cities. Variables that showed no significant difference were sex; travelling time to health centre and having visited a private doctor, practitioner of alternative medicine or faith healer during the previous year (Table 1).
Chronic self-reported illness was related with higher PCC. There were significant differences in 24 of the 33 categories. A positive correlation was seen between age, number of chronic illnesses and number of visits to GP, whereas fewer persons in the household were associated with higher rates of utilization (Table 2).
|
Table 2 shows that a lower score for social support (both confidant support and affective support) and family dysfunction were associated with a higher PCC. The highest positive correlation coefficient was found for the GHQ-28 score.
A multilevel analysis was developed. The null model with three levels showed that 93.29% of the variability was explained by the patients, 1.56% by GPs and 5.15% by health centres. We tried a model with only two levels, patients and health centres, because the variability explained by the GPs was very small. This model did not differ from the three-level model (chi-square = 0.83; P = 0.362). The two-level model with random effects was significantly different from the two-level model with fixed effects (chi-square = 28.11; P < 0.001). Hence, we selected a two-level model (patients and health centres) with random effects. We introduced into this model the independent variables at the patient level. No specific variable was introduced at the health centre level because we did not aim to explore differences between health centres. The variability explained (R2) was 0.48; 0.36 at the patient level and 0.12 at the health centre level. When we controlled for the confounding effect of the independent variables with multilevel analysis, the variable with the strongest association with higher consultations was mental health (GHQ-28 score), followed by family dysfunction. The number of chronic illness, urine infection, stomach problems, satisfaction and age were also associated with the number of PCC, whereas social support lost significance in the multivariate analysis. We found no first-degree interactions with P < 0.15 (Table 3).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Strengths and limitations of this study
The prospective design of our study enables us to affirm that psychosocial factors were determined prior to the outcome variable. We controlled for most patient variables that may influence PCC using a multilevel analysis, in accordance with current suggestions.29,30 We used a comprehensive approach to study the patients’ psychosocial factors because mental health, family dysfunction and social support are all factors that could interact and overlap each other,14 and also because, from a practical approach, management of mental health may be more effective if we also attend to family and social aspects. We also decided to measure PCC as a continuous variable; this approach avoided potential classification errors between ‘frequent attenders’ and control patients,31 as well as solving problems of comparisons derived from the different definitions and classifications.16 Our large, heterogeneous sample contributed to improving the external validity of the study. This study, therefore, might be compared (with due caution) with others undertaken from the perspective of a national health system in the context of urban populations, although it cannot be considered comparable with studies of private or fee-paying systems.
Although no significant differences were found in the distribution of sex and age between patients who were unavailable for interview and those who were interviewed, we cannot rule out a possible bias attributable to other variables that were not considered here. The rate of dropouts at the end of the follow-up period was reasonably small (7.3%), and practically no differences were seen between those who finished the study and those who did not. However, if we had not excluded these patients from the final analysis, the PCC could have been skewed (underestimated). Due to budget limitations, we chose just 11 health centres. This was a small sample at the health centre level and the influence of health centre level on PCC might therefore have been overestimated.
Professional and organizational factors can explain over 50% of variability of PCC.32 These factors are numerous,33 and include GP’s personal characteristics (age–sex, experience, type of training, values, orientations, attitudes, etc.), GP’s style of practice, practice setting (group style, peer pressure, organizational climate, role of clinical leadership, multiprofessional organization, repeat prescription organization, etc.), accessibility, continuity of professionals, interface with secondary care, list size, incentives and penalties. These groups of variables were not measured because we did not aim to study these factors. We were interested in the users’ psychosocial factors, and from this perspective the other factors and levels were considered as confounding factors and/or effect modifiers.
We did not assess the number of consultations with specialists, primary care nurses or private doctors during the follow-up because that information was unavailable; this might therefore have involved an ‘off-utilization’ bias. For example, some patients with chronic illnesses may go more frequently to hospitals and specialists and less so to GPs, although there is generally a moderate, positive correlation between hospital, specialist and GP visits.34,35 The patients who went to private doctors, as seen from our data before the follow-up, had slightly more visits with their GPs than those who did not go, though the difference was not significant. Additionally, some chronic patients may visit more often primary care nurses than GPs; however, positive and significant correlations also exist between both types of visits in Spain.36 Thus, despite we did not rule out the presence of an off-utilization bias, their effect was small. Although we did not examine the proportion of visits not registered in the clinical record, a previous study showed that the proportion was small.36
The prevalence of chronic illness was determined by patient interview rather than by review of medical records. The latter approach would have had the advantage of obtaining the information objectively. However, at the same time it would have had the drawback of not recording an illness felt but not expressed by the patient or expressed but not diagnosed or not taken into account by the doctor (e.g. mental health problems or minor health problems). On the other hand, morbidity ratings obtained from interviews with lists tend to be overestimated by frequent attenders. This may have skewed the results towards excessive weight being given to morbidity as an explanatory factor for PCC.37 Only two self-reported chronic illnesses were included in the final regression. This does not imply that no other chronic illnesses were associated with PCC. When the variables ‘age’ and ‘number of chronic illnesses’ were introduced in the model, most of the self-reported chronic illnesses lost their significance. The number of chronic illnesses is a variable that contributes to measuring ‘the need factor’, which is one of the most predictive factors explaining PCC. The ‘type of chronic illness’ is another factor, and each factor may have common and different aspects; for example, concerning the quantity or quality of the illness information. Nevertheless, we did not assess other variables related with the need factor (e.g. functional status, incapacity, quality of life, perception of illness susceptibility and severity).
The variable ‘satisfaction with your doctor’ was measured with only one item; hence, we neither attempted an approach to the multidimension concept of satisfaction nor did we aim to study it in depth. Satisfaction is a predisposing factor of PCC and we included it only as a factor to control for.
Standardized interview schedules were not used to measure mental health. However, in many countries the GHQ-28 is a questionnaire that has shown high correlation coefficients with standardized interview schedules, as well as possessing good reliability and validity coefficients.22
Comparison with existing literature
Our findings agree with previous cross-sectional studies that found an association between different psychiatric disorders and PCC.2–4 The few studies which used a prospective design also reported similar results.7–9 However, only a few studies have measured patient psychosocial factors from a comprehensive approach. In one of these15 mental health, family dysfunction and perceived susceptibility–severity were variables that were strongly associated with frequent attendance; however, this study was cross-sectional in design. Fiscella11 found similar results: higher perceived criticism (an aspect of family dysfunction) predicted more psychosocial/somatic and biomedical visits through self-rated mental health, whereas social support played no role in explaining these relationships. Sheehan38 related somatization, depression and poor social support with frequent attendance, although family dysfunction was not studied. The relationship between family dysfunction and PCC has an explanation from the family's conceptual framework, where behaviours are learned, attitudes are formed and patterns of stress adaptation are developed, including patterns of utilization.39 The most important source of social support in our culture is the family. This might explain the preponderance of APGAR score on the Duke-UNC scale in the final regression; moreover, both questionnaires had items that could overlap. In our sample, the low rates for social class, divorce and separation may also have influenced these results, and must alert us to the possibility of other cultural differences across populations which clearly could be relevant in interpreting issues of family dysfunction and social support. The patients’ exposure to the same life events, for example divorce, could be interpreted, valued and dealt with in different ways (in quality and intensity) in UK and in Spain; consequently, the weight of social support and family function as a cause of PCC could change between countries and cultures. From this point of view, the variability explained by family dysfunction might perhaps be higher in Spain than in UK. Further studies exploring the association of PCC with family and social factors, in different social and cultural contexts, are therefore needed.
Implications
PCC might best be understood using a multifactorial explanatory model including patients, doctors and organizational-related factors from prospective designs and with a multilevel analysis. There is a need to expand the explanatory models of PCC, and the role of psychosocial factors may be the key.40 Patient psychosocial factors should be studied with the inclusion of mental health, family dysfunction and social support, together with other related variables, such as health beliefs, illness behaviour or patient expectations.41 Our results might provide insights about why measures such as those described by Jiwa42 have no impact on frequent attenders, given that the intervention in that study was GP disease focused, while the causal factors identified in our study are patient–psychosocial related. From this practical perspective, our findings suggest that the effective management of mental health problems from a family-based approach may reduce PCC.
Summary of main findings and conclusions
Patient psychosocial factors were strongly associated with PCC, especially mental health and family dysfunction, whereas social support lost significance in the multivariate analysis. Chronic illness seemed less relevant in our study, and only two predisposing factors were finally included in the equation: age and satisfaction with the physician. The findings of this prospective study suggest that poor mental health and family dysfunction are the most important patient psychosocial factors predicting primary care utilization, whereas social support is of secondary importance. More comprehensive identification of psychosocial factors may enhance our understanding of PCC.
|
Declaration |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Funding: Spanish Medical Research Council; `Fondo de Investigaciones Sanitarias’ (FIS 930681); received a national award from the Spanish Society of Family and Community Medicine, `SEMFyC'.
Ethical approval: The study was approved by Granada Ethics Committee.
Conflicts of interest: None.
|
Acknowledgments |
|---|
We thank the Ministerio de Sanidad, Instituto Carlos III, grupo SAMSERAP and redIAPP (RD06/0018/0039) and the Sociedad Española de Medicina Familiar y Comunitaria. We thank the GPs and patients who participated in this study and Ian Johnstone for English language advice.
|
Notes |
|---|
Bellón JA, Delgado-Sánchez A, de Dios Luna J and Lardelli-Claret P. Patient psychosocial factors and primary care consultation: a cohort study. Family Practice 2007; Pages 1–8 of 8.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
1 Gill D, Sharpe M. Frequent consulters in general practice: a systematic review of studies of prevalence, associations and outcome. J Psychosom Res (1999) 47:115–130.[CrossRef][Web of Science][Medline]
2 Tweed DL, Goering P, Lin E, Williams JI. Psychiatric morbidity and physician visits: lessons from Ontario. Med Care (1998) 36:473–485.
3 Dowrick CF, Bellón JA, Gómez MJ. GP frequent attendance in Liverpool and Granada: the impact of depressive symptoms. Br J Gen Pract (2000) 50:361–365.[Web of Science][Medline]
4 Jones GN, Ames SC, Jeffries SK, Scarinci IC, Brantley PJ. Utilization of medical services and quality of life among low-income patients with generalized anxiety disorder attending primary care clinics. Int J Psychiatry Med (2001) 31:183–198.[CrossRef][Web of Science][Medline]
5 Ten Have M, Oldehinkel A, Vollebergh W, Ormel J. Does neuroticism explain variations in care service use for mental health problems in the general population? Results from the Netherlands Mental Health Survey and Incidence Study (NEMESIS). Soc Psychiatry Psychiatr Epidemiol (2005) 40:425–431.[CrossRef][Web of Science][Medline]
6 Bijl RV, Ravelli A. Psychiatric morbidity, service use, and need for care in the general population: results of The Netherlands Mental Health Survey and Incidence Study. Am J Public Health (2000) 90:602–607.
7 Vedsted P, Fink P, Olesen F, Munk-Jorgensen P. Psychological distress as a predictor of frequent attendance in family practice: a cohort study. Psychosomatics (2001) 42:416–422.
8 Ronalds C, Kapur N, Stone K, Webb S, Tomenson B, Creed F. Determinants of consultation rate in patients with anxiety and depressive disorders in primary care. Fam Pract (2002) 19:23–28.
9 Herrman H, Patrick DL, Diehr P, et al. Longitudinal investigation of depression outcomes in primary care in six countries: the LIDO study. Functional status, health service use and treatment of people with depressive symptoms. Psychol Med (2002) 32:889–902.[CrossRef][Web of Science][Medline]
10 Gomez Rodriguez E, Moreno Raymundo P, Hernandez Monsalve M, Gervas J. Socio-economic status, chronic morbidity and health services utilization by families. Fam Pract (1996) 13:382–385.
11 Fiscella K, Franks P, Shields CG. Perceived family criticism and primary care utilization: psychosocial and biomedical pathways. Fam Process (1997) 36:25–41.[CrossRef][Web of Science][Medline]
12 Penning MJ. Health, social support, and the utilization of health services among older adults. J Gerontol (1995) 50B:S330–S339.[Web of Science]
13 Kouzis AC, Eaton WW. Absence of social networks, social support and health services utilization. Psychol Med (1998) 28:1301–1310.[CrossRef][Web of Science][Medline]
14 Franks P, Campbell TL, Shields CG. Social relationships and health: the relative roles of family functioning and social support. Soc Sci Med (1992) 34:779–788.[CrossRef][Web of Science][Medline]
15 Bellón JA, Delgado A, Luna JD, Lardelli P. Psychosocial and health belief variables associated with frequent attendance in primary care. Psychol Med (1999) 29:1347–1357.[CrossRef][Web of Science][Medline]
16 Vedsted P, Christensen MB. Frequent attenders in general practice care: a literature review with special reference to methodological considerations. Public Health (2005) 119:118–137.[CrossRef][Web of Science][Medline]
17 McConnachie A, Wilson P, Thomson H, Ross S, Watson R, Muirhead P. Modelling consultation rates in infancy: influence of maternal and infant characteristic, feeding type and consultation history. Br J Gen Pract (2004) 54:598–603.[Web of Science][Medline]
18 Snijders TAB, Bosker RJ. Multilevel Analysis. An Introduction to Basic and Advanced Multilevel Modelling (1999) London: Sage Publications.
19 Instituto Nacional de Estadística. Censo Poblacional (1991) Madrid.
20 Alonso J, Pérez P, Sáez M, Murillo C. Validez de la ocupación como indicador de la clase social, según la clasificación del British Registrar General. Gac Sanit (1997) 11:205–213.[Medline]
21 Ministerio de Sanidad y Consumo. Encuesta Nacional de Salud (1989) Madrid.
22 Lobo A, Pérez-Echevarría MJ, Artal J. Validity of the scaled version of the General Health Questionnaire (GHQ-28) in a spanish population. Psychol Med (1986) 16:136.
23 Smilkenstein G, Ashworth C, Montano D. Validity and reliability of the family APGAR as a test of family function. J Fam Pract (1982) 15:303–311.[Web of Science][Medline]
24 Bellón JA, Delgado A, Luna JD, Lardelli P. Validez y fiabilidad del cuestionario de función familiar APGAR familiar. Aten Primaria (1996) 18:289–296.[Medline]
25 Broadhead WE, Gehlbach SH, Degruy FV, Kaplan BH. The DUKE-UNC functional social support questionnaire: measurement of social support in family medicine patients. Med Care (1988) 26:709–723.[Web of Science][Medline]
26 Bellón JA, Delgado A, Luna JD, Lardelli P. Validez y fiabilidad del cuestionario de apoyo social funcional DUKE-UNC-11. Aten Primaria (1996) 18:153–163.[Medline]
27 Tukey JW. Exploratory Data Analysis (1977) Reading, MA: Addison-Wesley.
28 Greenland S. Modeling variables selection in epidemiologic analysis. Am J Public Health (1989) 79:340–349.
29 Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav (1995) 36:1–10.[CrossRef][Web of Science][Medline]
30 Phillips KA, Morrison KR, Andersen R, Aday LA. Understanding the context of healthcare utilization: assessing environmental and provider-related variables in the behavioral model of utilization. Health Serv Res (1998) 33:571–596.[Web of Science][Medline]
31 Ekholm A. A review of methods for misclassified categorical data in epidemiology. Stat Med (1992) 11:271–275.[CrossRef][Web of Science][Medline]
32 Hurley J, Woodward C, Brown J. Changing patterns of physician services utilization in Ontario, Canada, and their relation to physician, practice, and market-area characteristics. Med Care Res Rev (1996) 53:179–206.
33 Eisenberg JM. Physician utilization: the state of research about physicians’ practice patterns. Med Care (2002) 40:1016–1035.[CrossRef][Web of Science][Medline]
34 Forrest CB, Nutting P, Werner JJ, Starfield B, Von Schrader S, Rohde C. Managed health plan effects on the specialty referral process: results from the Ambulatory Sentinel Practice Network referral study. Med Care (2003) 41:242–253.[CrossRef][Web of Science][Medline]
35 Gill JM, Mainous AG III. The role of provider continuity in preventing hospitalization. Arch Fam Med (1998) 7:352–357.
36 Bellón JA, Delgado A, Luna JD, Lardelli P. Influencia de la edad y el sexo en los distintos tipos de utilización en Atención Primaria. Gac Sanit (1995) 9:343–353.[Medline]
37 Connelly JE, Smith GR, Philbridk JT, Kaise DL. Healthy patients who perceive poor health and their use of primary care services. J Gen Intern Med (1991) 6:47–51.[Web of Science][Medline]
38 Sheehan B, Bass C, Briggs R, Jacoby R. Somatization among older primary care attenders. Psychol Med (2003) 33:867–877.[CrossRef][Web of Science][Medline]
39 Weimer SR, Hatcher C, Gould E. Family characteristics in high and low health care utilization. Gen Hosp Psychiatry (1983) 5:1983.
40 Bradley EH, McGraw SA, Curry L, Buckser A, King KL, Kasl SV. Expanding the Andersen model: the role of psychosocial factors in long term care use. Health Serv Res (2002) 37:1221–1242.[CrossRef][Web of Science][Medline]
41 Hodgson P, Smith P, Brown T, Dowrick C. Stories from frequent attenders: a qualitative study in primary care. Ann Fam Med (2005) 3:318–323.
42 Jiwa M. Frequent attenders in general practice: an attempt to reduce attendance. Fam Pract (2000) 17:248–251.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||