Assessment of pre-test probability in Primary Health Care using the International Classification of Primary Care-2 ( ICPC-2 )

Como citar: Gusso GDF, Lotufo P, Benseñor IM. Assessment of pre-test probability in Primary Health Care using the International Classification of Primary Care 2 (ICPC-2). Rev Bras Med Fam Comunidade. 2013;8(27):112-20. Disponível em: http://dx.doi.org/10.5712/rbmfc8(27)713 Palavras-chave: Atenção Primária à Saúde Medicina de Família e Comunidade Classificação Cuidado Periódico


Introduction
There is a long tradition in high-income countries of evaluation and research in Primary Health Care (PHC).For instance, in the 1950s the British general practitioners (GPs) started to analyse their daily work 1 and, in 1958, a study conducted by the Research Committee of the College of General Practitioners (which included 11 GPs) concluded that in around 50% of patient visits they could reach a diagnosis.In his classic 1963 article, Crombie describes similar results 2,3 .This trend has inspired family doctors and GPs to better code their work activity and, in the 1970s, those family doctors and general practitioners interested in taxonomy began to develop a classification system designed to be used in PHC.After some pilot studies that started as 'Reason for Encounter Classification' (RFEC), researchers have managed to launch the first version of the International Classification of Primary Care (ICPC-1).Although, Brazil was chosen as a site for piloting the ICPC-1, few studies in Brazil have used ICPC and none has assessed pre-probabilities in Brazilian PHC.
There are basically two modalities for assessing the complex and longitudinal relation between doctors and patients in PHC: the Encounter Modality (EM) and the Episode of Care Modality (ECM).In the EM every consultation is seen as a unique event, whereas in the ECM the "health problem or disease" is followed up "from its first presentation to a health care provider until the completion of the last encounter for the same health problem or disease" 4 .It is not the same as disease or illness episode, since disease and illness may continue to evolve after the last encounter with the health provider.Thus, the ECM as an analytical tool can better assess the continuity and the process of care.Nevertheless, it may be possible to apply the EM if the Reason For Encounter (RFE) is confronted against problems stated by doctors (that is, their diagnosis).Based on this latter methodological approach, this research has applied a quite simple technology (paper-based forms that can also be replicated in low-income areas worldwide), as well as the second version of the International Classification of Primary Care (ICPC-2), published by World Organization of Family Doctors (WONCA), to asses the RFE in Brazilian PHC.More specifically, the objective of this research was to evaluate the RFE of patients visiting the health centres, their diagnosis, the main comorbidities, and the pre-test probabilities for common diseases registered by family doctors in Florianopolis, Brazil.

Methods
Florianópolis is the capital of Santa Catarina, one of the states in the Southern region of Brazil.According to the 2000 National Census, 342,315 people lived there and 96.7% were literate 5 .Florianopolis adopts the Family Health Strategy for organizing its primary health care services, which entails that for an average 3.500 people the local health authority should provide a Family Health Team (FHT) that encompasses: one physician, one nurse, one or two nurse assistants and 6 health community agents.

Palabras clave:
Atención Primaria de la Salud Medicina Familiar y Comunitaria Clasificación Episodio de Atención Periódica In June of 2007 there were 48 health centres and 90 FHTs in Florianópolis, all of them with one generalist or family doctor 6,7 .Some health centres had computerized charts (Electronic Medical Record), and every doctor of the Family Health Strategy (FHS) had to classify each consultation using the ICD-10 choosing up to two codes per encounter, even for those working on paper-based record.This was a local health authority initiative, and it is worth mentioning that it is not required for the country as whole.
The data collection took place from June 21 st 2007 to June 20 th 2008.All 90 general practitioners or family doctors from the Family Health Strategy of Florianópolis were invited (21 personally and 69 by letter) and 30 agreed to participate: 15 personally and 15 by letter.
This cross-sectional study was designed in an Encounter Modality (EM), because it would be difficult to link encounters to episodes of care in a longitudinal fashion with paper-based forms.By using the EM this research has focused on two main questions: firstly, what are the reasons for encounter in the words of the patient?And secondly, what is the diagnosis (or problems)?This strategy enabled the calculation of pre-test probability for common diseases using the reason for encounter (the Appendix shows the research paper-based form used in the study).
All patients must have been registered in each period of research and, in the case of doctors facing a busy day, at least questions regarding reasons for encounter and problems should be stated.Reasons for encounter must reflect the patient´s words, while the problems must be defined in medical terms (i.e., the doctor should be assertive rather than framing it in terms of hypothesis or 'suspicious case').It was allowed that a 'diagnostic hypothesis' could be registered in the patient´s official record, but not in the research paper-based form.If it was impossible for the doctor to reach a diagnosis, the recommendation was to repeat the patient's signs or symptoms (as diagnosis) in the research paper-based form.Thus, the reasons for encounter on the left-hand side column had to be related to medically defined problems on the right-hand side column (see Appendix).If there were more than two reasons for encounter related to one problem, the two most important reasons should be selected.

Data collection
Each general practitioner answered the form in a typical week of work during each season (winter, spring, summer and autumn).The research paper-based form had nine questions regarding age, gender, marital status, occupation, type of consultation (booked or not), reasons for encounter (RFE), problems or diagnosis indicated by the doctor, interventions (plans, references and exams) and medication (new, chronic use or both).Each doctor could choose the day for data collection; however, at the end of each season, the information from all weekdays should be available for analysis.For example, Table 1 shows a typical week of work of a doctor: this doctor, in each research season, should choose one period of the working week, either morning or afternoon, for collecting the data.Only Tuesday and Friday afternoons could not be investigated since there were no individual consultations planned for these periods.Following this research schedule example, all seasons and all consulting days would be investigated and could be distributed according to the doctor's availability for the research.
Guidance on how to fill in the form and the ICPC-2 scheme was sent with the research paper-based forms before the first season of research (winter, 2007).This guidance highlighted the two most important methodological aspects of the study: • To complete the form immediately after each patient visit to avoid loosing any consultation of the chosen period; • To label the problem (or diagnosis) just if a disease was really present; if it was only suspected case, the doctor should register just the name of the sign or symptom.
The data was analysed through SPSS (Statistical Package for the Social Sciences) for Windows.All International Classification of Primary Care 2 nd edition (ICPC-2) alphanumeric codes were transformed in numbers.All ICPC-2 numeric codes were typed in a SPSS version 13.0 spread sheet together with other consultation form items, which resulted in a template.The referrals to specialists and other health professionals were coded following the relation of the Electronic Medical Record (EMR) used in Florianópolis PHC.When more than one exam was marked, the combination was included in SPSS on request (for example: blood plus urine without culture).Each reason for encounter and problem was coded by the main researcher (who is member of the WONCA International Classification Committee), following the established rules for coding 4 using terminologies' tools 8,9 .
All forms were transcribed to SPSS 13.0 by a research assistant, including the form numbers (first "variable").This research was approved by the Ethic Committee of the Hospital das Clínicas of the Medicine Faculty of the University of São Paulo State under protocol number 0180/07 and was part of a doctoral project.The volunteers did not receive any incentive and the authors have no conflict of interest to declare.

Results
The average age of physicians was 33 years (5.7) and 58.3% of the participants had between 6 to 10 years since they have graduated in medicine.From the 30 volunteers who agreed to participate, four have not filled any form.An average of 22.5 sets of completed forms per season has been sent back.All forms were considered, even if they were partially filled (i.e., if the data collection had not encompassed the whole week).An exclusion criterion was adopted, applied to doctors who had not completed at least one form per patient in each research period.A total of 5,698 forms were completed during one year research period: 1601 (28,1%) -winter; 1585 (27,8%) -spring; 1306 (22,9%) -summer; 1206 (21,2%) -autumn.The distribution of people included in the study was similar to the age-distribution of the population living in Florianópolis, except for adolescents of both sexes and young adult men, who use to be less frequent users of health centres.The most frequent reasons for encounter were prevention, fever and medication request for cardiovascular problems (Table 2) and the most frequent problems (or diagnosis) were hypertension, no disease (when patient came for prevention, for example) and acute upper respiratory infection (Table 3).It was not observed great variability according to different seasons.However, cough and asthma were slightly more prevalent in spring and autumn.When the reason for encounter was fever, the most frequent problems (or diagnosis) were acute upper respiratory infection, tonsillitis and fever (Table 4).In the opposite path, the most frequent reasons for encounter associated to the acute upper respiratory infection problem (or diagnosis) were cough, fever and throat symptoms/ complaints (Table 5).
The most common referred speciality was ophthalmology, with almost three times more referrals than the second one, which was "emergency".No drug prescription has been provided in just 26,6% of all encounters.

Discussion
Morbidity studies [10][11][12] have shown that up to 97.5% of consultations detect three or less problems.The present study found that about 30 health problems corresponded to more than 50% of all consultations.These findings are in line with those available from studies in other countries with similar design and methodology 10,12 .This is relevant because policy health makers can use this information for Continuing Professional Development programs, which should be based on these more prevalent reasons for encounters and problems.In the data collected, there were 1,625 reasons for encounters and 1,475 problems per consultation, which are similar to the findings elsewhere 10,11 .
Sometimes problems (diagnosis) received the same labels as the reasons for encounter.This seems to be a characteristic of primary care settings studied, as stated by Crombie 2 .In this regard, Weed 13,14 in 1968 suggested that one should state a disease just when there is "evidence" of it.If there is no certainty about what disease it might be, then the correct attitude is to choose a symptom or sign, or even to repeat the term used to characterize the reason for encounter, while waiting for an exam confirmation or for time resolution of the reason for encounter, also called "permitted delay" (or watchful waiting) in family medicine principles 15,16 .
Concerning the family doctors' age, available data from the whole country shows that 38.6% of doctors were 30 to 39 years old 17 , which means that this study's volunteers were younger than the doctors' national average.This can be explained by the expansion of the Family Health Strategy in Florianópolis through three public professional selection processes from 2004 to 2008, which attracted young professionals.The main difference, however, was in the qualification of family doctors: 79.1% of volunteers did at least residence medical training in Family and Community Medicine, when compared with the 14.4% national average.
The majority of the problems were comparable with the lists in previous studies [10][11][12][13]18 . In his research there was no wrong coding which could be considered as a "ragbag" of rubrics (_99) amongst the frequent problems, such as 'hypertension in child from zero to four years old' or 'male with gynaecologic problems'.The combination of participants with good training reporting the reasons for encounter and the problems and only one coder with experience in ICPC-2 probably have helped the study to reach a good quality of data.This often is not the case when the participant is also responsible for coding (the coder), situation in which wrong coding might be frequent.On the other hand, coding in the work routine enables to gather a great number of encounters, which is important to assess the pre-test probability of most problems.The most used components of the ICPC-2 were the component one (signs and symptoms -for reason for encounter) and the component seven (diagnosis and diseases -for problems or diagnosis).Considering hypertension, obesity and tobacco abuse as risk factors, and not as "diseases", from the most frequent problems (representing 50.4% of encounters -Table 3), only in 39.6% a disease was diagnosed.
Concerning patients' referral to other services, this research has found some differences when compared with previous studies.In this sample it was found that in 13.2% of encounters the patient was referred to a specialist, whereas in Takeda et al 18 the referral rate was only 9%.Official data from the Informatics Department of the local health authority (Secretaria Municipal de Saúde de Florianópolis) revealed that, from July 2007 to June 2008, in 8.7% of encounters patients were referred to a specialist.The difference might be explained by some specialists who give professional support to health centres, such as psychiatrists, geriatricians and paediatricians, whose consultations do not need to be registered in the Electronic Medical Record; however, the participants it this research may have registered their consultation in the paper-based form.
Most of the referrals were for ophthalmologist (19.4% of all referrals), reflecting Brazilian politics that forbid optometrists to prescribe corrective lenses.Compared to our data, in Netherlands the "eye surgeon" was just the fourth specialist referred to, representing only 8.2% of all referrals 10 .
The amount of prescriptions (at least one medical prescription in 73.4% of all encounters) supports the ritual of modern medicine and the roles which patients and doctors perform.Thus, it seems that for both sides it is not comfortable when neither a referral nor a prescription is done.At least one medication was prescribed (33.8%) even when the reason for encounter was preventive medicine!This study has some limitations: the first is the difficulty to code some terms and concepts.The development of an optimised thesaurus and a universal standardization of terminology is a great challenge for the WONCA's International Classification Committee, considering all cultural differences and language barriers.The second limitation is the uncertainty whether each doctor states the right problem.The doctors' qualifications in clinical reasoning, a good guidance on how to complete the form (EMR or chart), and enough training for coding would help to reach data of good quality.These limitations, however, do not diminish the importance of this type of research, since this is considered the best methodological approach to establish pre-test probabilities in primary care settings.

Conclusion
This study contributed to the knowledge of reasons for encounter and health problems of Florianopolis' population.The 30 most frequent problems involved 13 different chapters of ICPC-2, with regular distribution amongst them (from 1 to 4 problems per chapter).If each organ or system belonged to one different specialist, it would be necessary to have at least 13 specialists in each health centre.Family medicine/general practice does not cover all medicine fields but it covers the most prevalent and unspecific health problems and reasons for encounters.Studies using this methodology represent a strong tool to guide health authorities to develop strategies for continuing multiprofessional development.No data based on episode mode can be found in the Brazilian context and this is the only study that used reason for encounter and problem-diagnosis in an interrelated way for assessing the pre-test probabilities.Many countries are implementing Electronic Medical Records in health centres but high quality data can also be collected on paper-based records in developing countries.In summary, the ICPC-2 as a classification system is a great contribution to transform any health centre in a research centre, even those in rural areas of low-income countries.

Table 1 .
Hypothetic schedule of a family doctor during one typical week of work.

Table 2 .
Frequent reasons for encounter in the sample (50.4% of all reasons).

Table 3 .
Frequent problems in sample (50.4% of all encounters).

Table 4 .
Most frequent problems when reason for encounter was fever.

Table 5 .
Most frequent reasons for encounter when problem was Acute Upper Respiratory Infection.