Acute undifferentiated fever in Binh Thuan province, Vietnam: imprecise clinical diagnosis and irrational pharmaco-therapy

Authors


Corresponding Author Peter J. De Vries, Division of Infectious Diseases, Tropical Medicine and AIDS, Academic Medical Center, F4-217, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands. Tel.: +31 20 5664380; Fax: +31 20 6972286; E-mail: p.j.devries@amc.uva.nl

Summary

Objectives  To describe the characteristics of patients consulting commune primary healthcare posts for acute undifferentiated fever not being malaria (AUF), and to explore the diagnostic and therapeutic responses of the healthcare workers.

Methods  All patients presenting with AUF at 12 commune health posts and one clinic at the provincial malaria station, Binh Thuan, a dengue endemic province in southern Vietnam, were included. Record forms were used to fill in patient and diseases characteristics, pre-referral treatment, signs and symptoms, provisional diagnosis and installed treatment, referral and final outcome.

Results  Two thousand ninety-six patients were included from April 2001 to March 2002. The median delay to attend the health posts was, 0.87 day for >5, 1.15 days for children aged 5–15 years and 1.41 days for adults (P < 0.001). Sixty-five per cent of patients took some measures before consulting the health post, of whom 82% applied self-medication and 69% took antibiotics. Pre-referral medication with antibiotics increased with age (RR 1.012 per year of age; 95% CI: 1.004–1.019). The diagnostic and therapeutic response of healthcare workers was very unspecific. The tourniquet test was inappropriately used as general discriminating test, not only for detecting dengue haemorrhagic fever. Empiric antibiotic therapy was installed in 77.2% of cases.

Conclusions  Management of uncomplicated fever, not being malaria, at the primary healthcare level in Vietnam is non-specific, dominated by searching signs of hemorrhagic dengue and empiric antibiotic treatment. This can probably be improved by better education.

Abstract

Objectifs  Décrire les caracteristiques des patients consultant des postes communautaires de soins de santé primaire pour des fièvres aiguës indifférenciées autres que la malaria et explorer le diagnostic et les interventions thérapeutiques des agents de la santé.

Méthodes  Tous les patients présentant une fièvre aiguë indifférenciée dans 12 postes communautaires de santé et une clinique ont été inclus dans l’étude effectuée dans la station provinciale de malaria de Binh Thuan, une province endémique pour la dengue, dans le sud du Vietnam. Les formulaires de records ont été utilises pour saisir les caracteristiques de la maladie du patient, le traitement avant que le patient ne soit référé, les signes et les symptômes, le diagnostic provisionnel et le traitement instauré, le résultat au moment ou le patient a été référé et après.

Résultats  2096 patients ont été inclus dans l’étude entre avril 2001 et mars 2002. Les délais médians pour atteindre le poste de santéétaient de 0,87 jour pour les moins de 5 ans, 1,15 jours pour ceux âgés de 5 à 15 ans et de 1,41 jours pour les adultes (p < 0,001). 65% des patients ont recouru à d'autres mesures avant de consulter le centre de santé. Parmi ceux-ci, 82% ont eu recours à l'automédication et 69% ont pris des antibiotiques. La prise d'antibiotiques avant que le patient ne soit référé augmentait avec l’âge (RR: 1,012 par année d’âge; IC95%: 1,004–1,019). Le diagnostic et l'intervention des agents de santéétaient totalement non spécifiques. Le test du tourniquet était utilise de façon inappropriée comme test de discrimination et pas uniquement pour la détection de la fièvre hémorragique de dengue. Une thérapie empirique aux antibiotiques était instaurée dans 77,2% des cas.

Conclusion  La prise en charge des fièvres non compliquées et non malariques au niveau des services de santé primaire au Vietnam est non spécifique, dominée par la recherche de signes de dengue hémorragique et par un traitement empirique par antibiotiques. Cette attitude pourrait être améliorée par une meilleure éducation.

Abstract

Objetivos  Describir las características de pacientes que consultan puestos comunitarios de atención primaria por fiebre aguda indiferenciada (FAI), y explorar el diagnóstico y la respuesta terapéutica de los trabajadores sanitarios.

Métodos  Se incluyeron todos los pacientes con FAI en doce puestos comunitarios de salud y un hospital en la estación provincial de malaria en Binh Thuan, provincia de Vietnam del Sur endémica para dengue. Se utilizaron las historias clínicas para completar las características de los pacientes y la enfermedad, la medicación previa, los signos y síntomas, el diagnóstico provisional y el tratamiento establecido, la derivación y la resolución final.

Resultados  Se incluyeron 2096 pacientes entre Abril 2001 y Marzo 2002. La media en el retraso de atención en los centros de salud: 0.87 para menores de cinco años, 1.15 para niños entre 5–15 años y 1.41 para adultos (p < 0.001). El 65% de los pacientes tomó las mismas medidas antes de consultar un centro de salud, de los cuales el 82% se automedicó y el 69% tomó antibióticos. La automedicación con antibióticos aumentaba con la edad (RR 1.012 por año por edad; 95% CI: 1.004–1.019). El diagnóstico y la respuesta terapéutica de los trabajadores sanitarios fue muy inespecífica. La prueba del torniquete se utilizaba de forma inapropiada, no solo para detectar dengue hemorrágico sino como un teste discriminatorio general. La terapia antibiótica empírica se dio en un 77.2% de los casos.

Conclusiones  El manejo de la fiebre no complicada, no debida a malaria, en los centros de atención primaria de Vietnam es inespecífica y está dominada por la búsqueda de signos de dengue hemorrágico y por el tratamiento empírico con antibióticos. Tal vez esta situación podría mejorarse mediante campañas de educación.

Introduction

Vietnam has been highly successful in bringing malaria under control during the last decade (Ettling 2002). Despite the rapid decline of malaria (Ettling 2002; Hung et al. 2002; Nam et al. 2005), fever remains a common reason for seeking help at communal health posts. Unlike malaria, for which microscopic confirmation has become standard practice at many health posts, laboratory diagnosis of other infectious diseases is lacking, and diagnosis and treatment are generally only based on signs and symptoms.

Self-medication has become very common among febrile subjects in Vietnam and many other developing countries, similar to what was observed for malaria before the large-scale introduction of early diagnosis and treatment of malaria (EDTM) (Boonstra et al. 2002; Deressa et al. 2003). Unguided use of antibiotics has many disadvantages, such as selection of drug resistant micro-organisms, adverse drug effects, drug interactions and increased health expenditure (Larsson et al. 2000; Okumura et al. 2002).

In the case of malaria, we recently showed that improving the public knowledge and offering early diagnosis and effective treatment diverted patients from self-treatment towards professional help (Giao et al. 2005). We wondered if this can be achieved for other fevers as well. This requires further study into the interaction between health-seeking behaviour and the provided care. Health-seeking behaviour is frequently investigated by techniques, which are common in social sciences or marketing research (Font et al. 2001; Giao et al. 2005). The response of the healthcare provider is less well studied (Halfvarsson et al. 2000; Guyatt & Snow 2004).

Here we describe the characteristics of patients with acute fever, not being malaria, who present to a public primary health post and explore patterns in the healthcare workers’ diagnostic and therapeutic response.

Methods

The study was performed in Binh Thuan province (Figure 1) in southern Vietnam, starting in April 2001. Binh Thuan has a population of approximately 1.1 million on an area of 7992 km2, wedged between the Truong Son forested mountains (alt. 1100–1642 m) in the west and the South Chinese sea in the east. The majority of the population lives in rural areas, with approximately 187,042 people in and around the capital, Phan Thiet. The majority of the population (88%) is of Kinh (Vietnamese) origin, the others belong to several ethnic minorities of varying population size (Cham, K'Ho, Hoa, Tay, Nung, Ra Glai, Ta Lop and Ma) often living in the more remote areas.

Figure 1.

 Administrative map of Binh Thuan indicating the communes (name and number) participating in the study. Between brackets the population in 2000.

Until recently, Binh Thuan was a relatively poor region, and especially the ethnic minorities were vulnerable. Over the last decade the provincial annual income per capita increased to US$278 in urban areas and US$230 in rural areas (the national income per capita is US$374).

Healthcare is provided by a provincial hospital in Phan Thiet, nine district hospitals and 115 commune posts for primary healthcare and disease control (further called health posts). In 2001 there were 483 medical doctors with a university degree (MD) employed in the province. The target of the national strategy to staff every health post with at least one MD has almost completely been achieved by postgraduate training at university level of medical officers (the so-called second degree doctors). (Source: Statistical Yearbook 2001 – Binh Thuan Statistics Office, Phan Thiet) Additional professional training after formal graduation is virtually non-existent let alone ‘continuous education’.

This study took place in changing circumstances of liberalization of the health sector giving rise to the development of a private sector (Sepehri et al. 2003). To attend a health post patients pay a small fee (±US$0.06) and patients have to pay for examinations and drugs. Poor subjects, including members of the ethnic minority groups, are exempted from payment. The motives why patients choose the public or private sector are largely unknown.

Twelve, not adjacent, health posts and one clinic at the provincial malaria station, where febrile patients, suspecting malaria, come for diagnosis and treatment, were selected in a manner that would ensure a representative selection of rural and (semi-)urban, lowland and highland communes, and the province's ethnic population structure. The staff of the participating health posts were composed of MDs and second degree doctors.

All patients presenting with acute undifferentiated fever (AUF) were included in this study. AUF was defined as any febrile illness of duration less than 14 days, confirmed by an axillary temperature ≥38.0 °C, without any indication for either severe systemic or organ specific disease. Malaria was excluded by microscopic examination of a thick blood smear.

Data collection

Record forms were filled in for all AUF patients recording patient identifiers (age, sex, occupation and address), history of recent exposure factors such as occupation, fresh water contact, visiting forests, duration of disease and invalidation, treatment taken, signs and symptoms at presentation, provisional diagnosis and prescribed treatment, referral and final outcome. Diagnoses such as ‘acute fever’ and ‘viral infection’ were all reclassified to ‘undifferentiated fever’.

Blood samples were collected for sero-diagnosis, results of which will be presented elsewhere. All included subjects were asked to come back after 2–4 weeks for re-assessment and collection of a second blood sample.

Data were entered by the attending healthcare worker at the first presentation of the patient. All health posts were monitored at monthly visits by the research team from Cho Ray Hospital, Ho Chi Minh City.

The study was approved by the Review Board of the Cho Ray Hospital. The study was explained and discussed in meetings with provincial authorities and staff of the health posts. All patients (or, for children, the parents or guardian) gave written informed consent.

Data analysis

Statistical analysis was done using SPSS (Version 11.5, SPSS Inc., Ill.) and S-Plus 2000 (release 2, Mathsoft Inc, MA). Frequencies and means or medians were calculated to describe background variables. The chi-square test and the median test were applied to assess the relation of variables such as time of presentation at primary health posts and age groups, time of presentation at primary health posts and previous treatment, and season and presumptive diagnosis. A logistic regression model was used to explore the variables that related to antibiotic use. Associations between the different indicators were sought with explorative techniques such as correspondence analysis, an explorative cluster analysis technique and classification trees (Benzécri 1992; Venables & Ripley 1999).

Results

From April 2001 until April 2002, 2096 patients with undifferentiated fever (867 females and 1229 males, female/male = 1/1.3) were included. The median age was 18 years (range from 1 to 82). Their main occupation was farming (820 adults, 39%) and school attendance (768 children, 37%). Other occupations included construction and industrial labour (n = 153; 7%), civil officer (n = 39; 2%), child at home (n = 163; 8%) and retirement (n = 41; 2%).

Patients were divided into three age groups: adults >15 years (n = 1198; 57.2%), children from 5 to 15 years (n = 730; 34.8%) and >5 (n = 166; 7.9%). The age distribution of patients differed among health posts and is shown in Table 1. Some health posts preferentially attracted children (health posts 25, 51 and 103), whereas others were mainly visited by elderly subjects (e.g. Nos. 79 and 97).

Table 1.   Distribution of age of patients with acute undifferentiated fever who presented to 13 primary health posts in Binh Thuan province, Vietnam
Health post (code)Total patients (No.)Age groups (years)
<5 (%)5–15 (%)>15 (%)
  1. * Two cases missing age.

 7 451*7.319.772.9
11756.733.360.0
2521022.951.925.2
373151.345.153.7
49734.120.575.3
517724.754.520.8
591157.027.066.1
781677.225.167.7
79488.391.7
831613.146.650.3
9520912.438.349.3
979212.088.0
1031013.064.432.7
Total20947.934.957.2

Patients attended the health post relatively soon after the first symptoms. The median interval from onset of illness and of fever to presentation is shown in Table 2. The significant difference in time between the three groups suggests that parents seek help earlier for their children >5 than for their children between 5 and 15 years or themselves.

Table 2.   Patient delay in undifferentiated fever, presented to public primary healthcare posts in southern Vietnam, by age group and by referral pattern
 Patient delay in days (median, 90th percentile)χ2(d.f. = 2)*P-value
Age groups (years)
<55–15>15
  1. d.f., degrees of freedom.

  2. * Calculated by the Median Test.

From first symptom1.43 (3.95)1.64 (3.79)2.28 (4.89)61.03<0.001
From onset of fever0.87 (2.95)1.15 (3.13)1.41 (3.87)26.22<0.001
Help seeking behaviour
 Came directly to health postConsulted private clinic firstAfter self-treatment  
From first symptom2.44 (5.40)3.51 (7.52)2.42 (4.69)28.95<0.001
From onset of fever1.77 (4.71)2.68 (6.60)1.64 (3.77)42.47<0.001

Before attending the health posts, 1356 of 2096 (65%) patients took some measure such as self-medication (82%) or consulted a private clinic (11%) or a health post (5%). The latter group mainly comprises patients who did not meet the inclusion criterion of axillary temperature >38 °C at the first visit and who were enrolled at a later visit when fever persisted or recurred. Seeking healthcare at private clinics, but not applying self-medication, increased the patient delay (Table 2). Of these 1356 patients, 938 patients (69%) had already used antibiotics. Antibiotic use was lower among those who applied self-treatment (65%), compared with those who received treatment from a private clinic (94%) or public health post (83%; P < 0.001).

A logistic regression model was used to explore the relation between antibiotic use before attending the health posts and age, gender, season and health post. The relative risk (RR, 95% CI) of using antibiotics increased slightly by a factor 1.012 (1.004–1.019; P = 0.003) per year of age. The rainy season was also significantly associated with higher use of antibiotics (RR 2.310, 95% CI 1.820–2.933; P < 0.001). Neither gender nor health post appeared to be associated with antibiotic use. The main signs and symptoms of the patients are shown in Table 3.

Table 3.   Symptoms and signs of patients presenting with undifferentiated fever to public primary healthcare posts in southern Vietnam
 AllFrequency (%)
Age groups (years)
<55–15>15
Symptoms
 Sore throat89.546.488.296.3
 Anorexia80.481.976.882.4
 Myalgia46.87.229.762.7
 Headache45.833.753.043.2
 Running nose35.848.836.433.6
 Cough35.75628.437.4
 Nausea23.611.424.125.0
 Backache21.20.67.532.4
 Arthralgia16.32.48.822.8
 Vomitus11.717.513.29.9
 Abdominal pain10.49.08.911.4
 Rash2.90.61.83.8
 Haemorrhage1.10.62.50.4
Signs
 Pharyngitis56.064.567.547.8
 Myalgia38.46.624.051.7
 Rhinitis35.65328.237.6
 Pallor17.810.815.220.3
 Arthralgia7.72.45.110.0
 Dehydration7.19.04.88.2
 Conjunctivitis6.715.17.05.4
 Tender liver2.90.62.13.7
 Dermal Rash2.30.61.43.1
 Lymphadenopathy2.12.42.51.8
 Lymphadenitis1.70.62.91.2
 Jaundice1.600.82.3
 Bruise1.62.42.11.2
 Hepatomegaly1.00.60.71.3
 Splenomegaly0.500.30.8

The results of the tourniquet test, presented as the number of petechiae per square inch, are not shown in this Table. Petechiae appeared in 210 cases (10%; ≤9/square inch: n = 126; 10–19/square inch: n = 50; ≥20/square inch: n = 34), and among all age groups [>5: n = 20 (9.5%), 5–15 years: n = 127 (60.5%) and >15 years: n = 63 (30%)].

The response of the healthcare worker was assessed in two ways, viz. the presumptive diagnosis and the treatment given. The frequency of the different presumptive diagnoses is shown in Table 4. Diagnoses varied by season at all age groups. There were no significant differences between males and females with respect to presumptive diagnosis.

Table 4.   Primary healthcare workers’ presumptive diagnosis of patients with undifferentiated fever, by age and season
 Frequency of presumptive diagnoses (%)
Age group (years)*
<5†5–15‡>15§
Season
Dry, n = 79Rainy, n = 87Dry, n = 330Rainy, n = 400Dry, n = 599Rainy, n = 599
  1. Undiff. fever, undifferentiated fever.

  2. * Two cases missing age.

  3. Chi-square Test:

  4. † χ2 (4 d.f.) 12.23, P = 0.016.

  5. ‡ χ2 (5 d.f.) 31.78, P < 0.0001.

  6. § χ2 (8 d.f.) 17.84, P = 0.022.

  7. ¶ Others: Allergy (8), Gastritis (7), Lymphadenitis (5), Arthritis (2), Mumps (2), Clinical malaria (2), Measles (1) and Varicella (1).

Undiff. fever (n = 1074)34.243.736.144.058.360.9
Pharyngitis (n = 500)27.827.629.427.324.417.0
Dengue fever (n = 180)5.114.910.016.54.56.2
Tonsillitis (n = 147)24.18.020.37.52.21.8
Typhoid fever (n = 75)1.301.22.03.76.7
Diarrhoea/enteritis (n = 58)6.33.41.21.03.53.5
Leptospirosis (n = 11)00000.71.2
Hepatitis (n = 10)000.30.50.30.8
Others¶ (n = 39)1.32.31.51.32.51.8

The diagnostic process of the healthcare worker is visualized in the classification tree in Figure 2. In this model, diagnostic leads such as gender, age, season, exposure factors, and signs and symptoms were entered as independent variables and presumptive diagnosis was taken as dependent variable. In most cases the diagnosis was classified as ‘undifferentiated fever’ without an identifiable pattern of diagnostic leads. In 1968 cases the presumptive diagnosis was made more specific by using one or more diagnostic leads. In 962 of these, the tourniquet test appeared to be the most significant factor for differentiating between dengue and a group of other diagnoses. The regression tree analysis chooses the best cut-off value of the number of petechiae, observed with the tourniquet test. This was very low: 1.5 per square inch. Other leads used were the presence of a red pharynx on physical examination, the complaint of diarrhoea, myalgia, cough or abdominal pain and finding abdominal tenderness on physical examination. However, many cases were still misclassified, including less common diagnoses such as typhoid, tonsillitis, leptospirosis and hepatitis.

Figure 2.

 Classification of primary healthcare workers’ strategies to make a presumptive diagnosis of patients with undifferentiated fever, based on signs and symptoms. Classification tree of the presumptive diagnosis of 1968 of 2098 patients, based on their signs and symptoms. This tree is not a clinical algorithm to guide diagnosis but an analytical tool for post hoc inference on the diagnostic strategy used by the health workers. It identifies the relation between signs and symptoms and the presumptive diagnosis for fever patients. The tree shows all decision nodes, i.e. the signs and symptoms used for making the presumptive diagnosis. Final nodes are depicted by rectangles. The ratio in each node indicates the proportion of incorrect classifications. For example, if initially the presumptive diagnosis of all 1968 patients would be classified as ‘undifferentiated fever’, the most common diagnosis, then 962 would be incorrect. The first and most important discriminating sign used by the healthcare workers is the number of petechiae occurring in the tourniquet test. Below a cut-off value of 1.5/square inch, the best possible classification would be 1768 cases of undifferentiated fever (of whom 785 would be incorrect); above the cut-off 200 cases would be classified as dengue (of whom 53 incorrect). At the bottom of the tree the total proportion of incorrect classifications has decreased from an initial 962/1968 to 668/1968.

Forty-six healthcare workers contributed to this study. Their prescribed treatment was not uniform except that antipyretics appeared in 98.3% (range 78–10%) of prescriptions. Vitamins were prescribed in 87.2% (39.3–100%) of cases, antibiotics in 68.7% (16.7–100%), corticosteroids in 9.6% (0–33.1%), IV fluids in 20% (0–83.3%), anti-tussants in 13.5% (0–46.5%) and oral rehydration solution (ORS) in 10.3% (0–82.6%). The distribution of prescriptions is shown in Table 5, divided by age groups. There were significant differences among age groups except for the prescription of corticosteroids. Notably, no patient left the health post without a prescription.

Table 5.   Distribution of treatment for patients
TreatmentAll age n (%)Age groups (years)χ2, (d.f. = 2)P-value
<55–15>15
n (%)n (%)n (%)
  1. ORS, oral rehydration solution; d.f., degrees of freedom.

Antipiretic1976 (94.3)159 (95.8)705 (96.6)1110 (92.7)13.6730.001
Vitamin1683 (80.3)110 (66.3)621 (85.1)950 (79.3)31.8940.000
Antibiotic1524 (72.7)122 (73.5)581 (79.6)819 (68.4)28.8470.000
Fluid359 (17.1)7 (4.2)101 (13.8)251 (21.0)37.3800.000
Cough medicine347 (16.6)34 (20.5)100 (13.7)213 (17.8)7.4590.024
Corticoid236 (11.3)21 (12.7)77 (10.5)137 (11.4)0.7280.695
ORS160 (7.6)28 (16.9)54 (7.4)78 (6.5)22.2540.000

Table 6 shows the data of presumptive diagnoses and treatments. It was recognized that antipyretics, vitamin, antibiotics and fluid were prescribed for all of the presumptive diagnoses but anti-tussant and corticoids were used more frequently for a presumptive diagnosis of pharyngitis and tonsillitis and ORS was used mostly for diarrhoea and dengue fever. Combinations of different antibiotics were found in 82 cases (5.4%). The most frequently prescribed antibiotics were amoxicillin (43.2%) and cephalexin (36%). In general, low dosages of antibiotics were applied for short intervals. When these regimens were compared with internationally used recommendations, e.g. the ‘Stanford Guide to Antimicrobial Therapy’; 2003, it appeared that the dose was appropriate in 88.6% of patients but the duration was appropriate in only 54.4%. Both appropriate dose as well as treatment duration was prescribed in 47.3%.

Table 6.   Presumptive diagnosis and prescribed treatment in patients presenting to primary health posts
Presumptive diagnosis (No.)Frequency of prescriptions (%)
AntipyreticVitaminAntibioticIV fluidsAnti-tussantCorticoidORS (%)
  1. ORS, oral rehydration solution; Undiff. fever, undifferentiated fever.

  2. * See Table 4.

Undiff. fever (1075)97.186.658.618.211.04.24.7
Pharyngitis (501)93.078.099.211.028.321.83.6
Dengue fever (180)97.886.751.130.03.31.131.1
Tonsillitis (147)92.563.3100.06.148.346.32.7
Typhoid fever (75)92.070.793.313.310.72.72.7
Diarrhoea (58)67.234.570.741.40.0048.3
Leptospirosis (11)90.972.7100.072.70.000.0
Hepatitis (10)60.040.070.020.00.010.00.0
Others (39)*76.969.274.42.65.123.12.6

Table 6 does not show the interdependency between the multiple outcomes. With correspondence analysis this can be illustrated in a comprehensive manner. Correspondence analysis is a rather old technique, which receives more attention lately (Benzécri 1992). In correspondence analysis the relative frequency of all outcomes is calculated and expressed in terms of co-ordinates on scales of virtual dimensions. These co-ordinates are calculated in such a way that on a two-dimensional plot the correspondence of different outcomes is expressed as the distance from the origin (dimension 1 = dimension 2 = 0) in the same direction. The correspondence between presumptive diagnosis and prescribed treatment is shown in Figure 3. For example, the relative frequency of diarrhoea corresponds with that of ORS, indicating that healthcare workers preferably prescribe ORS for diarrhoea.

Figure 3.

 Correspondence between presumptive diagnosis and prescribed treatment. On two virtual scales of relative frequency, the correspondence between the presumptive diagnosis (triangles) of febrile patients and the prescribed treatment (round dots) by primary healthcare workers are shown. The association between presumptive diagnosis and treatment can be read from the plot as departures from the origin (dimension 1 = dimension 2 = 0) in the same direction. Thus, for example, the preferred treatment for diarrhoea is with oral rehydration solution (ORS). No, no prescription.

Almost all patients returned home after consultation, except 73 (3.5%) patients who were admitted to the health post and 33 (1.6%) who were referred to a higher healthcare level.

Discussion

This study shows that the case management of undifferentiated fever at the public primary healthcare level in Vietnam is characterized by very short patient delay, high rates of prior self-treatment, unspecific diagnostic considerations and poly-pharmacy with high prescription rates of antibiotics.

The short patient delay is probably the result of the rigid malaria control policy in this province, which stimulates febrile patients to come to a health post immediately, where they can receive microscopic diagnosis and adequate treatment at low cost or free of charge. This feature is rather unique to this region (Giao et al. 2005). Other studies in developing countries have shown that patient delay may be much longer (Khe et al. 2002).

Patient delay in children was shorter than in adults. This is a very common feature. Parents are probably more alarmed about the health of their children than they are about themselves. This has been found in other studies also (Larsson et al. 2000; Giao et al. 2005). However, some recall bias may play a role here, since there may be a time span between onset of fever and the moment on which the parent or guardian notices that a child, especially an infant who cannot talk, is ill.

Self-medication is very common in countries where drugs can be obtained over the counter (Kamat & Nichter 1998; Deressa et al. 2003; Nam et al. 2005). In Vietnam the effectiveness of national and provincial malaria control programmes is probably the reason why febrile patients seek help at public health posts. But even in that context it is apparently a widespread habit to apply self-medication at the first sign or symptom.

The signs and symptoms most frequently point at the upper respiratory tract as the focus of infection, with sore throat and pharyngitis being the most frequent complaint and finding, respectively. Tonsillitis in children and pharyngitis in adults are mainly diagnosed during the dry season whereas in children dengue fever is preferentially diagnosed in the wet season. The data do not provide an explanation for the seasonal effect. Seasonal difference in health seeking behaviour or the healthcare workers’ perception of pre-test likelihoods may play a role. Real seasonal differences in transmission of upper respiratory tract pathogens in the tropics show a predilection for the rainy season albeit less distinct as in the winter of colder climates (Shek & Lee 2003).

The healthcare workers’ unspecific diagnostic response is not unique to this province. Many studies of the diagnostic considerations at the primary healthcare level show similar findings (Halfvarsson et al. 2000; Phillips-Howard et al. 2003; Sepehri et al. 2003). Of all the diagnostic leads, the tourniquet test was interpreted by the health workers as the most discriminating. The tourniquet test is an old test. It aims to measure the hemorrhagic status of a patient, which is often the result of thrombocytopenia, thrombopathy and vasculopathy. In modern medicine this test is hardly used anymore, mainly because other indicators are available such as the platelet count and newer tests of haemostasis. For dengue however, where vasculopathy and thrombocytopenia may co-exist, the tourniquet test still has a place in the diagnostic classification (World Health Organization 1997). The healthcare workers in Binh Thuan, similar to many other dengue endemic regions, have been trained to use the tourniquet test. However, they do not only use it for assessment of severity of dengue, they apparently also use it for separating dengue as a diagnosis from other fevers and intuitively apply a lower cut-off point for the number of petechiae per square inch. This interpretation of the tourniquet test is unlikely to increase the diagnostic accuracy.

The therapeutic response of the healthcare workers, i.e. prescribing drugs for all and antibiotics for most patients, is extreme but not unique (Sepehri et al. 2003; Yanagisawa et al. 2004). Although the diagnosis ‘undifferentiated fever’ might exclude bacterial infections, this apparently was no reason to withhold antibiotics. Especially the use of corticosteroids seems an inadequate response to a short-lived fever. However, it is common knowledge that the patient's subjective improvement after the use of corticosteroids is impressive, and that for a healthcare worker, not restrained by much knowledge of evidence-based medicine, it is a logical step to improve the patients well-being as soon as possible.

This study indicates some potential points for improvement in the management of undifferentiated fever. First, febrile patients or parents of febrile patients seek help shortly after onset of symptoms. The patient delay may be even too short as most of the diseases are self-limiting by nature. The low threshold for early detection of malaria is probably one of the main reasons for this short patient delay and high consumption of healthcare. In order to reduce the workload of the public primary healthcare service, a two-step policy, with screening for malaria for all, but further consultation only for severe cases, based on carefully defined criteria, can be considered. However, the patients’ perceptions of fever and disease and economic incentives should be taken into consideration. For example, a high workload can also mean a high income for a healthcare worker.

Second, the presumptive diagnosis of the healthcare worker can be improved. The most rational approach would be to upgrade the educational level of all healthcare workers. As mentioned previously, such an approach is currently being carried out by the Vietnamese government. In addition, at every health post the epidemiology of the locally prevalent infectious diseases should also be known. These data are often available, but not used as a way to increase the prior likelihood of the presumptive diagnosis. This should be part of, ongoing, postgraduate training. Another way of improving the diagnosis, by providing rapid confirmatory tests, is a point for further study.

Last, improved, rational, pharmacotherapy may do much benefit. The drawbacks of unguided drug use are evident: high costs, potential side effects, selection of resistant micro-organisms and so on. This requires extra-education, including postgraduate training, of the healthcare workers. It definitely also needs some re-education of the population, to redress its hunger for drugs.

In conclusion, management of uncomplicated fever at the primary healthcare level in Vietnam can be improved by better specification of the diagnosis, better knowledge of local diseases and more rational pharmacotherapy. These objectives can be achieved by better, postgraduate, education.

Acknowledgements

The study was supported by the Dutch Foundation for the Advancement of Tropical Research (WOTRO). We gratefully acknowledge the contributions of the healthcare workers at 13 studied sites. We would like to thank Prof. Truong Van Viet, MD, PhD, the director of Cho Ray Hospital, HCMC and the authorities of Binh Thuan province for their cooperation.

Ancillary