Family Health Program and Ambulatory Care-sensitive Conditions in Southern Brazil

OBJECTIVE: Ambulatory care-sensitive conditions (ACSC) are health problems managed by actions at the fi rst level of care. The need for hospitalization by these causes is avoidable through an effective and proper primary health care. The objective of the study was to estimate ACSC among patients hospitalized by the Sistema Único de Saúde (Brazilian Health System). METHODS: Hospital-based cross-sectional study involving 1,200 inhabitants of Bagé (Southern Brazil) who were inpatients between September/2006 and January/2007. The patients answered a questionnaire applied by interviewers and were classifi ed according to the model of attention utilized prior to hospitalization. ACSC were defi ned in a workshop promoted by the Ministry of Health. The variables analyzed included demographic and socioeconomic characteristics, health and health services utilized. Multivariate analysis was conducted by the Poisson model, according to a hierarchical conceptual framework, stratifi ed by sex and model of care. RESULTS: ACSC accounted for 42.6% of the hospitalizations. The probability that the main diagnosis for hospitalization is considered an ACSC is greater among women, children under fi ve years of age, individuals with less then fi ve years of schooling, hospitalization in the year prior to the interview, emergency room consultation, and being an inpatient at the university hospital. Among women, ACSC are associated with age, educational level, length of time the health center has been in existence, living in an area covered by the Programa Saúde da Família (Family Health Program), use of this service, emergency room consultation during the month prior to the interview and hospital to which patient was admitted. For men, it was associated with age, have undergone another hospitalization in the year prior to the interview and hospital to which patient was admitted. CONCLUSIONS: Analysis of ACSC allows identifying groups with inadequate access to primary health care. Although we could not infer an effect on the risk of hospital admission, analysis by sex and model of care suggests that Family Health Program is more equitable than " traditional " primary health care.


INTRODUCTION
Primary health care is the model of care that receives most endorsement from the World Health Organization.Its goal is to improve health indicators, reduce the gaps in morbi-mortality, and achieve a more rational consumption of biomedical technology, thus attaining greater effi ciency in expenditure within this sector.a In Brazil, the Sistema Único de Saúde (SUS) [Unifi ed Health System] and the Programa Saúde da Família (PSF) [Family Health Program (FHP)] follow the same conceptual foundations, and primary health care is thus the point of departure in an attempt to redirect the model of assistance.
Evaluation and monitoring of actions and results are fundamental in adapting the policies and the actions implemented.The Sistema de Informação da Atenção Básica [Data System on Primary Care], the Programa de Expansão e Consolidação da Saúde da Família [Program of Expansion and Consolidation of Family Health], 13 and the Pacto dos Indicadores da Atenção Básica [Pact of Primary Care Indicators] 19 as well as the bills concerning fi nancial aid to research are cited among the governmental efforts accomplished in this direction.However, instruments and studies evaluating the impact of primary care on the population's health are lacking. 15The Data System on Primary Care is useful in organizing the family health teams, 1 but it presents limitations insofar as evaluative research is concerned.Furthermore, it is not used to register data on "traditional" primary care (that is, not FHP), which impedes comparisons between the two models of attention.The evaluation of the Program of Expansion and Consolidation of Family Health is restricted to municipalities with over one hundred thousand inhabitants.The Pact of Primary Care Indicators lacks an aggregate measure that would permit a more integrated evaluation of primary care.
On the other hand, hospitalizations due to Ambulatory Care Sensitive Conditions (ACSC) have expanded as an indirect indicator of access to opportune and effective care at the primary level of health care.b,c ACSC are health problems that are attended by actions characteristic of primary health care.Hospitalizations for ACSC are an indirect indicator of the effectiveness of the health system, at this level of care.It is presumed that individuals hospitalized for ACSC did not receive effective health care at an opportune moment, leading to an aggravation of their clinical condition, which then required hospitalization. 9,10is indicator was created at the end of the 1980s in the United States to evaluate the impact of lack of access a Starfi eld B. Atenção primária: equilíbrio entre necessidades de saúde, serviços e tecnologia.Brasília: Unesco/Ministério da Saúde; 2002.b Mendes EV.A atenção primária à saúde no SUS.Fortaleza: Escola de Saúde Pública do Ceará; 2002.c Fundação João Pinheiro.Centro de Estudos Econômicos e Sociais.Atenção básica à saúde em Minas Gerais: desigualdades na distribuição de recursos fi nanceiros e na prestação de serviços básicos após a introdução do Piso da Atenção Básica (PAB) [  Feb 20].Available at: http://tabnet.datasus.gov.br to primary health care services. 5,6,20Afterwards, it was employed in countries with universal access, such as Spain, to compare the effectiveness of different models of primary health care. 4,8In Brazil, studies have been conducted with this indicator in the States of Ceara (Northeastern Brazil) and Minas Gerais (Southeastern Brazil).b According to the equity principle, the implementation of the Family Health Program is initiated in areas where poverty is greater, whose populations also suffer more occurrences of diseases and limitations with respect to access to services, including health care.Thus, the socioeconomic characteristics of the population must be taken into consideration when interpreting the effect of the model of attention on hospitalizations for ACSC, particularly when coverage by the FHP is not universal.When such variables are lacking, the aggregation of individuals' conditions may serve as a proxy.
ACSC make it possible to identify, in an objective and comparable manner, parts of the population that lack adequate primary care, which is presumed to be convenient for the Unifi ed Health System in Brazil.Thus, this study seeks to answer the following question: compared to other models utilized in organizing care, the FHP is associated to a smaller proportion of ACSC among the hospitalized population?The objective of the present study was to estimate the probability of the diagnosis of ACSC among residents of a municipality, hospitalized by SUS, according to the model of care utilized in the consultations prior to hospital admission.

METHODS
The population of the municipality of Bage (Southern Brazil) was estimated in 122,461 inhabitants in 2006, 82% of which were living in the urban zone.It has three hospitals (one of which is an army hospital that has no covenant with SUS), 470 hospital beds in covenant with SUS (one hospital bed for each 3.8 thousand inhabitants) and 21 unidades básicas de saúde (UBS) [primary health care centers (PHCC)], 13 of which are Family Health Centers that attend 52% of the population.d All admissions to hospitals in covenant with SUS of residents of the municipality between the 16th of September 2006 and the 15th of January, 2007 were analyzed.The patients (or their guardians, when they were children) were interviewed during the period of hospitalization, by trained interviewers, during all days of the week.
The questionnaire included demographic and socioeconomic data as well as information concerning the use of health services.The diagnosis justifying admission and discharge from the hospital were collected in hospital records, after the patient was discharged.
Obstetrical admissions, admissions that terminated in death, patients transferred to the Intensive Care Unit (ICU) of other hospitals and people who were not capable of answering the questionnaire, according to the interviewer's evaluation, as well as those who refused to reply, were excluded.Patients admitted to the ICU were interviewed after being transferred to a room.No age limit was established for inclusion in the study.
The minimum number of hospitalizations necessary in order to detect a difference of at least 10% of the probability of an ACSC diagnostic among the hospitalized patients, according to whether he or she resided in an area covered by the Family Health Program was 868 individuals, taking the arcoseno approximation 17 with parameters alfa=5%, beta=20%, losses=10%, global proportion of ACSC estimated in 46%, FHP: not -FHP of 1:1.Another 30% was added in order to confer greater stability to the adjustment for confusion factors, and the resulting sample was calculated in at least 1.129 subjects.
Data was digitalized and partially processed through the EpiData Entry program.
Evaluation of the PHCC was undertaken in a discussion group with the Director of the Health Department and the team coordinating the municipality's primary health care.The criteria utilized in this evaluation were: commitment with the individuals utilizing health services, team work, intersectionality, integrality, equity, communitarian work and the organization of care.At the end of the discussion, each PHCC received a score from zero to ten with respect to each of the criteria.
The outcome being studied is the proportion of ACSC among the hospitalizations analyzed.The list of codes of the 10 th revision of the International Classifi cation of Diseases (ICD-10) considered ACSC was defi ned in a workshop promoted by the Ministry of Health, in December, 2005 a (Table 1).During this process, inspired by Caminal et al's study, 11 a group of experts answered questions with respect to the frequency, transcendence, sensitivity to primary health care actions, ease of diagnosis, need for hospitalization in response to disease development, and the infl uence of fi nancial incentives on the codes registered, among a list of pre-selected causes.The outcome variable, created with the aid of the EpiData Analysis program, considered the ICD recorded at diagnosis upon discharge from the hospital.A person was considered FHP user if he/she was a resident of the region covered by the Family Health Program and if he/she had been attended at a consultation at this unit during the previous month (whether or not the motive of this consultation was related to the cause that led to hospital admission).
After describing the data and conducting bivariate analysis, a multivariate analysis was conducted through the Poisson model according to a hierarchical theoretical framework for determining hospitalization for ACSC (Table 2).Within this model, determinants of health conditions and use of the health services may be found in its most distal level, at the intermediate level, descriptors of health conditions and use of the services prior to hospital admission can be found and, at the proximal level, are the hospital services utilized as well as the performance of the municipality's primary health care services.
All the variables at each level of determination being analyzed, starting from the most distal level to the outcome, were included in the equation and those with p≥0.2 were eliminated step-by-step.Each subsequent level incorporates the variables previously maintained; the measures of effect and the p-value reported for each variable are those found in its level of analysis.Multivariate analysis was conducted for the entire set of individuals and for each sex strata as well as for each model of care.Except when indicated, the p-values presented were obtained by means of Fisher's exact test or, when comparing averages, by Student's t-test or ANOVA; the Stata 9.0 program was utilized in analysis.
Participants signed a Term of Informed Consent before responding to the questionnaire.The study was approved by the Committee on Ethics in Research of the UFPel [Universidade Federal de Pelotas, Rio Grande do Sul State].

RESULTS
Between 16/Sept/2006 and 15/Jan/2007 1,446 non-obstetric admissions of municipal residents that qualifi ed for reimbursement by SUS were recorded in the hospitals of Bage.These individuals were approached by the study team and the proportion of refusals to participate in the research was 6.6% (96 cases), the average age of the latter was 19.8 years older that those interviewed (p<0.001), with no signifi cant difference with respect to their sex.In 47 cases, (3.5%) the interviewee informed that he/she resided in another municipality, being excluded from the study; these cases did not differ from those of participants according to hospital of admission, age or sex.Among the 1,303 subjects that agreed to participate in the study, 103 (7.9%) were hospitalized more than once, thus a total of 1,200 cases of single hospital admissions were analyzed.The variables with the greatest proportion of missing data were: diagnosis at discharge (6.0%), family income (7.4%),another hospitalization during the previous year (8.4%) and health plan's consultation as the source of recommendation for hospital admission (5.4%).
The interviewees had an average of 5.7 (SD=4.0)years of schooling and lived in the urban zone (96.2%), with a mean family income (during the month prior to the      **** The variable was entered into the model without the isolated criteria of evaluation, so as to avoid the effect of collinearity. the analysis of the diagnosis reported at the time of admission to the hospital.The mean family income was R$ 69.82 lower for those individuals hospitalized for ACSC than those admitted to the hospital for other causes (p=0.007); the mean level of education for both groups was similar (there was a difference of 0.3 years; p=0.2).

Table 3 presents the relation between the model of care and ACSC diagnosis upon admission to the hospital.
There was no signifi cant difference according to the model of care or use of the health care center of reference.However, independently of the population being analyzed, the proportion of ACSC was greater among individuals who indicated a FHP Centre as his/her service of reference or when the patient had consulted with a physician at his primary health care center during the previous month.
Bivariate analysis indicated a greater probability that diagnosis at admission to the hospital would be ACSC among: women, children under 5 yrs of age, people who had less schooling, those who were poorer, those who had been hospitalized at least one other time in the year prior to the interview, and among those who had been to a consultation, during the month before the interview, for a reason other than the one that led to hospitalization.The hospital in which the patient was admitted as well as the specialty of the admitting physician, were also associated with the outcome.Although referrals from health plan consultations recommending that patients be admitted to the hospital reached p=0.2, there was a small number of observations of this type (n=37) and the variable was not included in the multivariate model.
Table 4 presents the description and bivariate analysis of the variables that entered the adjustment model, as well as the multivariate analysis, involving all study subjects and according to sex strata.It may be observed that receiving primary health care within the FHP, the length of time the primary health care center (PHCC) has been in existence and, principally, the period of time it has been functioning as FHP were variables, close to the limits of statistical signifi cance.Among the performance criteria evaluated by the Director of the Health Department and the team coordinating the municipality's primary health care, the one which was closest to an association with ACSC as a cause of hospitalization was "organization of care" (p=0,1).
As it may be observed on Table 4, the proportion of ACSC among women was 28% greater than among men.The age group with the greatest risk of fi nding an ACSC among the patients admitted to the hospitals through SUS was composed of children under fi ve years of age and this risk increased once again among adults aged 50 or older.The proportion of ACSC decreased with increasing educational level (p<0.001).The same table indicates that when the FHP was the service of reference, the probability of a diagnosis of ACSC was 15% lower (p=0.08).The hospital was the main associated factor.Patients whose PHCC of reference received an above average score for the criteria "team work" presented a probability 19% lower than others that the cause of admission to the hospital was an ACSC (p=0.06).The probability of ACSC diagnosis increased for: the female sex, age under 5 years, less than fi ve years of schooling, hospitalization in the previous year, consultation at the emergency room, admission at the university hospital.
Furthermore, Table 4 indicates that the major differences between the sexes refer to the model of attention.The proportion of ACSC was 33% lower among women residing in a FHP region, but 32% higher among those that had had a FHP consultation in the month prior to the interview.Among men, no association was observed between the model of attention and ACSC.Among women, ACSC prevalence was also associated to: age group, schooling, length of time the PHCC was functioning, having a Family Health Center as the center of reference, being a person who utilizes the FHP services, having consulted at the emergency room of a hospital during the month prior to the interview, and hospital to which she was admitted.For men, ACSC prevalence was associated to age group, having been hospitalized at least one other time the previous year and hospital to which he was admitted.
By analyzing separately the model of care of reference (Table 5), certain singularities are perceived: sex is only associated to one ACSC in the FHP and in traditional primary care; in the other models of care (consultation by specialist, at the hospital ambulatory, at the business or union, through a health plan, private physician, among others), the variable that had the strongest effect and statistical signifi cance was ownership of consumer goods.Among these models of care, no association was found with the hospital or physician responsible for admission.The same table indicates that the tendency to decrease the proportion of ACSC as years of schooling increases loses signifi cance among the group that is covered by the FHP.Compared to all other patients residing in Family Health Areas, those that had a consultation at the PHCC in the month prior to the interview had a 20% higher probability that the cause of admission to the hospital was an ACSC (p=0.05).A consultation at the emergency room and the physician's specialty were only associated with the outcome among patients residing in Family Health Areas.On the other hand, criteria for evaluating the performance of the health teams were only associated to the outcomes of "traditional" primary care.

DISCUSSION
This study is a hospital survey, which investigated the measure of the effect of the proportion of ACSC among hospital admissions.Inferences may not be made concerning the risk of hospitalization for these causes.A larger population covered by a particular model of care could result in a greater proportion of ACSC among hospitalized patients who use this model of care, even if the rate of hospitalizations among them was lower.
The associations found can only be interpreted as an effect of the model of care when referred to analysis within each model (as exemplifi ed by the proportion of ACSC according to the levels of education in each model of care), or when the analysis refers to the entire study population (such as the proportion of ACSC according to the hospital to which patients were admitted).This may be an explanation for the apparently paradoxical relation between the lower probability of ACSC among patients residing in a Family Health Area and FHP users.
Thus, whenever the factor of analysis requires the residing population as a denominator, or the population that consulted for an ACSC and was not admitted to the hospital (such as the proportion of ACSC according to the model of care), its effect could be estimated only if the hospitalized population was compared to the population which was not hospitalized.It is expected that there will be, among the study population, an over-representation of people exposed to greater risks and in worse health conditions than the general population.The variable FHP user, is useful in identifying the portion of the population covered by the FHP who did not use the system in the month prior to the interview, but perhaps incorporates aspects of the subject's health status, such as the gravity of the problem that led to the hospitalization, more than regular use of the FHP.This possibility is reinforced by the fi ndings of the Programa de Expansão e Consolidação da Saúde da Família-UFPel 13 [Program of Expansion and Consolidation of Family Health -Federal University of Pelotas] that indicate that the poorest among the poor utilized the PHCC to which they were assigned according to their residential zone.
Selection biases are not expected (beyond those that are due to seasonal and temporal variation), since all patients admitted to the hospitals during this period were interviewed, with only a small proportion of refusals, despite the fact that average age of this group was higher.It is not possible to dismiss eventual classifi cation errors, particularly when the diagnosis upon admission to the hospital is infl uenced by the systems of reimbursement.However, advances in auditing within SUS are notorious and the few studies that discuss the trustworthiness of diagnosis recorded on the Autorização de Internação Hospitalar (AIH) [Hospital Admission Authorization] form conclude that it is useful in epidemiological studies. 18Such a bias, if it occurs, should be non-directional since there is no reason to expect any difference due to the patient's PHCC of reference.
Hospitalization for ACSC may occur, among other reasons, due to the lack of use or a delay in the use of resources offered by the primary health care center as well as to inadequate clinical management.However, it should be considered that people seek solutions to their health problems and health professionals seek to work in an appropriate manner.As indicators of the effectiveness of the health system, under any circumstances, the interpretation of these hospitalizations should be directed towards the conditions of the organization of the system that generate them and not to the patients or the health workers.For example, the lack of knowledge concerning the supply of services and other "invisible" barriers to the established services perceived by the citizens or, on the other hand, the inadequate training of health professionals.
The ACSC do not evaluate personal determinants of the patient's clinical condition or the quality of the medical act that lead to hospitalization, but the result of policies and actions executed in the face of such health problems.As in this study, individual's variables may be used to identify compliance with the principles of equity of a particular model of care, or its inclusion in models of analysis of variables relative to the health system (as proxy of the living conditions of the population being cared for), but should not be utilized to infer the individual's responsibility concerning his need for hospitalization.
The lower probability of ACSC among the women that reported they lived in areas covered by the FHP did not occur among the men, a difference which may be due to the fact that women utilize primary health care services much more than men.The apparently paradoxical associations observed among the residents in areas of family health and among those that effectively attended consultations in the FHP during the month prior to the interview, besides expressing the lack of a populational indicator, may be due to the possible incorporation of clinical gravity in the variable FHP user.
It is plausible that the service where the individual made his/her consultation prior to hospital admission is the one with which he/she has the closest bond.However, the lack of knowledge concerning previous service utilization makes it impossible to state that this occurs in all cases, and this should be taken into consideration when interpreting the results.On the other hand, asking individuals to report on health care during a broader period of time could increase the memory bias.
When the model of care is SUS's primary care, the main characteristics of the person associated to avoidable hospitalization were sex, age and educational level of the patient.In the private sector and other forms of organization of care, the main variable associated was the possession of consumer goods, which indicated a probability twice as great, among subjects with worse economic conditions, that the cause of hospitalization would be an ACSC.This characteristic, which suggests a compensating effect of the Unifi ed Health System on social inequity, is even more evident in the FHP, where the association with educational level loses statistical signifi cance.The FHP attends more and recommends hospitalization less often than all other models of care.
In Brazil, accompanying the avoidable hospitalizations in order to evaluate and monitor primary health care was proposed by the State of Ceara; studies promoted by institutions associated to the State Health Departments were conducted in Bahia and Minas Gerais.a,b The Ministry of Health has also been studying this indicator (this resulted in the list of ACSC codes c utilized in this study), but no Brazilian study on the theme was found in scientifi c journals.To the authors' knowledge, this is also the fi rst Brazilian study in which the indicator is integrated, and which uses primary and not aggregated data.
The proportion of ACSC found (43%) is much greater than that reported in other studies, even among the Spanish and Australian (who have universal access and extensive lists of ACSC), that varied from 7% to 13% of total hospital admissions, [2][3][4]8,10,12 although there is little comparison possible between the populations and the health systems studied. In the panish example, with a system that is effectively universal, only municipalities with a single "Primary Health Care Area" were studied, which might explain such a low proportion of these cases.8 On the other hand, the proportion found in this study is similar to that found in Minas Gerais (where ACSC represented 33% of the total discharges, which included deliveries and abortions (20% of discharges), excluded in our study).a The other results of this study are consistent with those found in the literature, that indicate greater rates of hospitalization among populations with worse socioeconomic conditions, among age groups at both extremes of the spectrum and varying according to the organization and supply of services, without a great deal of consistency according to sex.4,6,7,16,b The lack of association to the residential zone (rural or urban) suggests that there are no barriers with respect to access to primary health care or hospitalization for these different populations of the municipality.3 Despite an incipient effect in the population studied, the FHP in Bage already presents signals of favorable impact.Socioeconomic discrimination was not observed among the individuals hospitalized, thus approaching the objective of the equity principle.The results observed among women and in the analysis by strata of the model of care suggest that even though it has been in activity for a short period of time, the FHP presents better results than traditional primary health care.However, the more immediate solutions for reducing the proportion of ACSC among hospital admissions in the municipality seem to be related to the organization of care in the university hospital.
Criteria for evaluation elaborated with the administrative team seem appropriate for traditional primary care, but need to be improved for FHP.However, this study represents only the perspective of this group and not of others, such as the patients and the health workers of the system.
This indicator is still new and its concept is being developed, 9 as may be attested by the large variety of lists of codes considered ACSC as well as the different spheres of its application.The effect of the hospital on the determination of ACSC may signify that there is a need to render this measure appropriate theoretically.In this sense, the complexity that results from the multiplicity of health problems accounted for within this indicator, with different probabilities of expression according to sex, age group and its relations to the health services, indicate a broad space for research, both theoretical and in the fi eld, involving the entire population, or specifi c age groups, sexes and causes, both on the ecological and the individual level of data analysis.

2
Health condition, use of services Other hospitalization(s) in the 12 months prior to the interview; other consultation(s) in the 30 days prior to the interview; length of time the FHP has been functioning; the individual utilizes the FHP; model of care in the consultations in the 30 days prior to the interview for the problem that generated the hospitalization; recommendation of hospitalization, in these consultations.3 Use of the services, service performance hospital to which the patient was admitted; specialty of the physician responsible for admission; evaluation of the services by the administrator.Outcome: Hospitalization for Ambulatory Care -Senstive Conditions CCEB-ABEP: Criteria of Economic Classifi cation Brazil, of the Associação Brasileira de Empresas de Pesquisa [Brazilian Association of Business Research]

Table 2 .
Hierarchical model of analysis of the study variables.Sex, age group, schooling, per capita family income, possession of consumer goods (CCEB-ABEP), residential zone (rural or urban), length of time residing in the current address.

Table 1 .
Brazilian List of Ambulatory Care-Sensitive Conditions.
Source: Ministry of Health.Department of Primary Care.Workshop for the elaboration of a Brazilian List of Hospitalization for Ambulatory Care-Sensitive Conditions.Belo Horizonte, MG, Dec/9/2005.

Table 3 .
Hospital prevalence of Ambulatory Care -Senstive Conditions and the use of health services.Bage, SouthernBrazil,  2006-2007.
* Poisson Model with robust variance FHP: Family Health Program ACSC: Ambulatory Care -Sensitive Conditions

Table 4 .
Hospitalizations for Ambulatory Care -Sensitive Conditions, by sex according to the main study variables.Description and multivariate analysis by levels of determination.Bage, SouthernBrazil, 2006-2007.

Table 5 .
Multivariate analysis of hospitalizations due to ambulatory care-senstive conditions by levels of determination, according to the model of care.Bage, Southern Brazil, 2006-2007.* Given an incidence of hospitalization for ACSC of approximately 2% per year, a population sample with this objective would have a very high budget.Identifi cation of adequate denominators would be possible if a variable of the PHCC of reference to the hospitalized patient were included in the Sistema de Informações Hospitalares/SUS [System of Hospital Information of the Unifi ed Health System] and if a record of the population residing in each health care area were available, as occurs in the FHP.