Knowing a medical doctor is associated with ... - Wiley Online Library

Tropical Medicine and International Health

doi:10.1111/j.1365-3156.2006.01744.x

volume 11 no 12 pp 1868–1877 december 2006

Knowing a medical doctor is associated with reduced mortality among sick children consulting a paediatric ward in Guinea-Bissau, West Africa M. Sodemann1,2, S. Biai2,3, M. S. Jakobsen2 and P. Aaby1,2 1 Projećto de Sau´de Bandim, Bissau, Guinea-Bissau 2 Danish Epidemiology Science Centre, Statens Serum Institut, Copenhagen, Denmark 3 Hospital Nacional Simaõ Mendes, Bissau, Guinea-Bissau

Summary

background To examine equity in access to public health services in Guinea-Bissau. methods The study was conducted in 2000–2001 at the emergency clinic of the only paediatric ward in Bissau. Mothers of all children from the study area were interviewed about previous care seeking and relations with anybody working in the health sector. All management actions in the emergency clinic were registered. In-hospital and subsequent community mortality was ascertained through community surveillance. The measured outcome was mortality risk within 30 days of first consultation. results We followed 1572 children with a first consultation. Of these, 8.2% died within 30 days. Acquaintance with a physician reduced 30-day mortality risk by 48% (95% CI: 18–66). The effect was strongest among post-neonatal children (54%; 95% CI: 18–74). Mortality within 30 days of consultation was also independently predicted by consultation after 7 pm, nurse team on duty, day of week and young mother. In a multivariate model, socioeconomic status and school education were not associated with 30-day mortality when acquaintance with a medical doctor was taken into account. conclusion Favouritism may be a significant factor for quality of care and child mortality in developing countries. Interventions to improve hospital and health worker performance should be given high priority. keywords child mortality, health systems research, equity, quality of care, hospital, community survey, Guinea-Bissau, Sub-Saharan Africa

Introduction During the past 10–15 years, most developing countries (DC) in sub-Saharan Africa have undergone extensive structural re-adjustments, which have incorporated a considerable cut in government health budgets. The countries have adopted a district and community-based health system with decentralization of resources and decision making. At the same time, vertical primary health care (PHC) programmes have absorbed a relatively large proportion of global funds for health improvements. This has left the referral level in the hospital sector in a financial vacuum with poor performance, low staff morale and low confidence as a result (Rutkove et al. 1990; Van Lerberghe et al. 1997). Equity and access to public health services in DC has become a priority for donors, e.g. The World Bank investing in the health sector. Most equity studies have focused on PHC. Studies on equity and quality of care in hospital settings have not been carried out. There are certain indications that better curative services are needed, as most PHC programmes have 1868

a limited effect on age groups with the highest mortality rates (Fauveau et al. 1990; Schofield & Ashworth 1996; Weber 2000). Furthermore, these interventions do not reach the poorest population groups (Castro-Leal et al. 2000; Victora et al. 2003). A small number of studies assessing hospital performance and problems of triage have pointed to the weak spots in the quality of care for sick children seen at the third level hospitals in low-income countries (Mirza et al. 1990; Sodemann et al. 1997; Reyes et al. 1998; Barreto et al. 2000; Nolan et al. 2001). Unfortunately, there are no studies of the influence of illegal or unofficial user fees, corruption or ‘favouritism’ on type of care provided at different levels of public health care (Mebtoul et al. 1999). In 1989, the Bandim Health Project (BHP) in GuineaBissau established a surveillance system linking the community registration system and the only paediatric ward in the country. Through this surveillance system, it has become increasingly clear that favouritism, defined as making use of an acquaintance or family ties with a person working in the health sector, could potentially affect child

ª 2006 Blackwell Publishing Ltd



M. Sodemann et al. Knowing a medical doctor

survival in settings with limited resources (Sodemann et al. 1997, 2004). In a previous study, we have shown that mothers are concerned with the importance of favouritism in relation to successful care seeking (Sodemann & Rodrigues 2005). We therefore carried out a study of child mortality in relation to care-seeking behaviour testing the hypothesis that maternal acquaintance with a health person would increase chances of survival among sick children consulting at the outpatient and emergency clinic of the paediatric ward, Simaõ Mendes National Hospital, Bissau, Guinea-Bissau.

Methods Hospital and community surveillance Data were collected at the Department of Paediatrics, Simaõ Mendes National Hospital and through the surveillance system of the BHP, Bissau, Guinea-Bissau. Since 1978, the core of the surveillance system has been a continuous follow-up of mothers and children through home visits to collect information on pregnancies, births, growth, vaccinations, childhood diseases, use of health services and survival. As part of the ongoing community studies, the Bandim Project has, since 1989, conducted a continuous registration of all hospitalizations from the study area at the (only) paediatric ward in Bissau. In 1997, registration was expanded to include all hospitalizations at the ward. A team of field assistants daily collect information on diagnoses and survival during hospital stay as well as status at discharge: ‘cured’, ‘improved’, ‘dead’ or ‘fled’. The ward has 100 beds and there are normally 6000–6500 admissions per year. Mothers were interviewed, and for children from the study area, information was collected to enable identification in the community register. The longitudinal community study covers around 15% of the hospital’s catchment population. It is possible to link hospital and community data before and after hospitalization. Information on socioeconomic class, background factors, birth history, previous hospitalizations, previous loss of children and survival was derived from the longitudinal demographic health surveillance system in Bandim (Sodemann et al. 2002). Study population From June 2000 to November 2001, two field assistants registered all consultations at the emergency clinic. Each mother bringing a child was interviewed upon arrival, by a trained field assistant, about previous care seeking, family or social relation with anybody working in the health sector. Information on ‘acquaintance or familiarity with


health-related personnel’ was asked as a simple question: ‘Do you know anybody working in the health sector or Ministry of Health?’ If the answer was positive, the respondent was asked to indicate the type of profession and place of employment of the person. Information on acquaintance with a health person was not obtained from mothers of 188/1572 (7.5%) enrolled children. The main reason for this lack of information was that in acute clinical situations parents were sometimes frightened, disturbed and stressed, in which case the field assistant had instructions not to interview them. All management actions in the emergency clinic were registered together with timing. Time to attention was measured from when the mother arrived at the clinic to first attention from either the nurse or the medical doctor on duty. Clinical indicators Because of the focus on favouritism, we applied several direct and indirect measures of clinical severity at the first visit: waiting time to attention, hospitalization and, as an external measure, the field assistant conducting the interview was asked to rate the clinical state of the child (severely ill, moderately ill and not very ill); this assistant has more than 10 years of interviewing at the paediatric ward. As many hospitalized children flee and these children may have an increased mortality risk, outcome was defined as the mortality risk within 30 days of the first visit at the outpatient clinic. Medical doctors were grouped into three case fatality groups (high, medium and low) according to the level of case fatality of the children they hospitalized during the study period. Medical doctors were asked to assess the level of clinical skills of their colleagues in three groups (inexperienced, somewhat experienced or quite experienced). This was considered the closest we could do to obtaining an objective validation of the level of medical experience because of the very diverse medical training Guinean doctors have received in terms of country they graduated from, quality of that medical training as well as post-graduation training often in more than 10 countries for each doctor. Medical doctors were grouped into one of two groups according to their tendency to hospitalize children from the outpatient clinic. Discharged children and children not hospitalized were followed with routine home visits of the BHP until December 2003. Statistical methods Risk ratios for mortality were estimated as odds ratios (OR) by logistic regression. Risk factors for 30-day mortality were first analysed in three models with 1869




appropriate adjustment of estimates: background factors (Table 1), care-related factors (Table 2) and factors associated with maternal encounter with the health system (Table 3). In the multiple regression analysis, predictors from each of these models were included in the final model with a significance level of 0.1 and eliminated by backward elimination with a significance level of 0.1. The presence of interaction was assessed by adding an interaction term together with adjustment terms with a significance level of 0.05. Trends in proportions were estimated with Mantel– Haenszel test. The population-attributable risk was defined

as PAR ¼ P · (MR ) 1)/[P · (MR ) 1) + 1], where P is the proportion exposed and MR the mortality risk.

Results Care-seeking pattern and mortality At the hospital, 1572 children were registered at their first consultation, 589 (37.5%) were hospitalized immediately and 23 children were hospitalized following repeated consultations within 30 days of the first contact. Within

Mortality risk, Odds ratio (OR, 95% CI) Crude OR Overall 129/1572 Sex Male 57/854 Female 72/718 Child age 0–30 days 30/83 31–364 days 35/436 1–2 years 42/512 3–4 years 9/218 5–15 years 13/323 Mother’s age 14–20 years 29/173 21–54 years 100/1399 Mother’s school education More than 7 years 34/468 4–7 years 35/525 1–3 years 23/247 None 37/332 Socioeconomic status Poorest 18/138 Less poor 68/881 Richest 42/551 Number of pregnancies 6–10 18/169 1–5 111/1376 Ethnic group Pepel 53/564 Manjaco 20/249 Muslim 11/154 Mancanha 11/181 Balanta 17/153 Mixto 6/95 Other 11/176 Residential area Bandim 1 62/770 Bandim 2 36/407 Belem 31/392 Mother gave birth in hospital Yes 61/779 No 61/687

Adjusted OR

(8.21)

–

(6.67) (10.0)

0.64 (0.47–0.92) 1

0.48 (0.31–0.73) 1

(36.1) (8.00) (8.20) (4.13) (4.02)

6.33 0.98 1 0.48 0.47

5.59 1.08 1 0.45 0.46

(16.8) (7.15)

2.62 (1.67–4.10) 1

2.23 (1.26–3.95) 1

(7.26) (6.67) (9.31) (11.1)

0.82 (0.47–1.42) 0.57 (0.35–0.92) 0.82 (0.47–1.41) 1

0.81 (0.43–1.53) 0.61 (0.35–1.07) 0.81 (0.43–1.53) 1

(13.0) (7.72) (7.62)

1.82 (1.01–3.26) 1.01 (0.68–1.51) 1

1.87 (1.00–3.82) 1.09 (0.69–1.72) 1

(10.7) (8.07)

1.25 (0.59–1.33) 1

0.91 (0.49–1.69) 1

(9.40) (8.03) (7.14) (6.08) (11.1) (6.32) (6.25)

1 0.84 0.74 0.62 1.21 0.65 0.64

1 1.33 0.77 0.50 1.27 0.64 0.69

(8.05) (8.85) (7.91)

1.02 (0.65–1.59) 1.13 (0.68–1.86) 1

1.13 (0.68–1.89) 1.18 (0.66–2.09) 1

(7.83) (8.88)

0.87 (0.60—1.26) 1

0.79 (0.50–1.24) 1

(3.66–11.0) (0.61–1.56) (0.23–1.00) (0.25–0.88)

(0.49–1.44) (0.38–1.46) (0.32–1.22) (0.68–2.14) (0.27–1.56) (0.32–1.26)

Table 1 Post-consultation 30-day mortality: background risk factors; 1572 sick children seeking consultation at the paediatric outpatient and emergency department of the Simaõ Mendes National Hospital, Bissau, Guinea-Bissau

(2.82–11.0) (0.63–1.84) (0.20–0.98) (0.23–0.90)

(0.71–2.47) (0.34–1.75) (0.22–1.10) (0.64–2.54) (0.22–1.88) (0.32–1.49)

Adjusted for age, hospitalization at the first visit and severity assessment of the assistant.

1870





Table 2 Post-consultation 30-day mortality: clinical history and care-related risk factors; 1572 sick children seeking consultation at the paediatric outpatient and emergency department of the Simaõ Mendes National Hospital, Bissau, Guinea-Bissau

Deaths/children in group (%) Consultation in First quarter 23/338 (6.80) Second quarter 24/233 (10.3) Third quarter 43/442 (9.73) Fourth quarter 39/559 (6.98) Severity rating Severely ill 52/145 (35.9) Moderately ill 59/663 (8.90) Not very ill 18/764 (2.36) Consulted anywhere 48 h, less than 14 days Yes 54/381 (14.2) No 75/1191 (6.30) Hospitalization within 30 days No 8/958 (0.84) At the first visit 115/590 (19.5) Only after one or more 6/24 (25.0) further visits Mother previously lost a child No 68/949 (7.17) Yes, once 40/421 (9.50) Yes, twice or more 18/154 (11.7) Child previously hospitalized Yes 13/149 (8.71) No 116/1423 (8.15) Parents sent out to find money and drugs Yes 37/244 (15.2) No 92/1328 (6.93) Waiting time to attention 0–25 min 29/177 (16.4) 26–180 min 60/636 (9.43) >180 min 36/742 (4.85)

Mortality risk Odds ratio (OR, 95% CI) Crude OR

Adjusted OR

0.64 (0.35–1.16) 1 0.94 (0.55–1.59) 0.65 (0.38–1.11)

0.81 (0.40–1.61) 1 1.13 (0.59–2.11) 0.85 (0.43–1.58)

29.5 (15.7–55.3) 5.89 (3.26–10.7) 1

3.87 (1.72–8.68) 1.09 (0.51–2.32) 1

0.98 (0.53–1.82) 1

0.46 (0.24–0.91) 1

2.46 (1.70–3.56) 1

2.16 (1.38–3.39) 1

1 28.8 (13.97–59.4) 39.6 (12.5–126)

1 19.5 (7.91–47.9) 40.6 (11.8–139)

1 1.36 (0.91–2.05) 1.72 (0.99–2.97)

1 1.38 (0.84–2.21) 1.36 (0.70–2.61)

1.08 (0.59–1.96) 1

0.86 (0.44–1.71)

2.40 (1.60–3.61) 1

0.72 (0.44–1.18) 1

3.84 (2.28–6.46) 2.04 (1.33–3.13) 1

1.69 (0.92–3.10) 1.15 (0.70–1.88) 1

Adjusted for age, hospitalization at the first visit and severity assessment of the assistant.

30 days of the first visit, we recorded 129 deaths in the cohort; 104 died in the hospital, 17 were hospitalized but died at home and 8 died at home without being hospitalized.

died, 24 had been born at the hospital, 2 at the health centre and 4 at home. Clinical and management factors

Background factors The relation between background factors and 30-day mortality is shown in Table 1. Socioeconomic status and school education had the expected relation to mortality in univariate analysis as did child age, maternal age and ethnic group. The lower mortality for boys was not explained by severity (P ¼ 0.42), immediate attention (P ¼ 0.53) or chance of hospitalization (P ¼ 0.36). Eightythree children were of neonatal age at the time of consultation, of whom 30 eventually died. Of those who


Factors related to clinical history, clinical state and management in relation to 30-day mortality are listed in Table 2. The severity rating of the field assistant correlated well with the 30-day mortality risk [OR ¼ 0.20 (0.15–0.27); for every step less severe, P-trend < 0.0001]. Adjusting for the factors in the final mortality model did not alter the significance level of the trend. The assistant identified 90.0% (530/589) of children who were hospitalized at the first visit with a rating of either moderately or severely ill. With the same rating, the assistant identified 1871




Deaths/children in group (%) Maternal acquaintance or familiarity with Medical doctor Yes 28/526 (5.32) No 83/858 (9.67) Nurse Yes 33/364 (9.07) No 78/1020 (7.65) Other health workerà Yes 12/120 (10.0) No 99/1264 (7.83) Field assistant Yes 8/130 (6.15) No 103/1254 (8.21) No information 18/188 (9.57) Clinical team on duty (period 1) 1 11/176 (6.25) 2 13/168 (7.74) 3 18/169 (10.7) 4 8/165 (4.85) 5 10/150 (6.67) Clinical team on duty (period 2) 1 19/221 (8.60) 2 20/131 (15.3) 3 13/143 (9.09) 4 10/138 (7.25) 5 7/111 (6.31) Tendency group of paediatrician to hospitalize Likely group 105/1180 (8.90) Less likely group 24/392 (6.12) Paediatrician mortality Highest 52/549 (9.47) Medium 57/641 (8.89) Lowest 20/382 (5.24) Paediatrician experience level Experienced 62/703 (8.82) Less experienced 67/869 (7.71) Mother arrived between 7 pm and 7 am Yes 36/253 (14.2) No 93/1319 (7.05) Consultation at day of week Monday 21/264 (7.95) Tuesday 25/266 (9.40) Wednesday 12/274 (4.38) Thursday 20/232 (8.62) Friday 26/225 (11.6) Saturday 12/171 (7.02) Sunday 13/140 (9.29)

Mortality risk Odds ratio (OR, 95% CI) Crude OR

Adjusted OR

0.51 (0.33–0.79) 1

0.55 (0.33–0.92) 1

1.20 (0.79–1.84) 1

1.28 (0.77–2.12) 1

1.31 (0.70–2.46) 1

1.04 (0.48–2.23) 1

0.73 (0.35–1.54) 1 –

1.62 (0.71–3.70) 1 –

0.56 0.70 1 0.42 0.60

(0.25–1.23) (0.33–1.49)

0.60 0.58 1 0.29 0.43

(0.24–1.49) (0.24–1.39)

1.40 2.68 1.49 1.16 1

(0.57–3.43) (1.10–6.59) (0.57–3.85) (0.43–3.15)

1.20 2.66 1.74 0.99 1

(0.41–3.47) (0.88–8.03) (0.55–5.05) (0.30–3.26)

(0.18–0.92) (0.27–1.34)

(0.10–0.80) (0.17–1.11)

1.50 (0.95–2.36) 1

1.27 (0.74–2.16) 1

1.89 (1.11–3.21) 1.76 (1.04–2.99) 1

1.91 (1.03–3.52) 2.16 (1.16–4.01) 1

1.16 (0.81–1.66) 1

1.31 (0.86–2.00) 1

2.19 (1.45–3.30) 1

1.83 (1.12–2.98) 1

0.66 0.79 0.35 0.72 1 0.57 0.78

0.88 0.68 0.40 0.86 1 0.40 0.88

(0.36–1.20) (0.44–1.41) (0.17–0.71) (0.39–1.33) (0.28–1.18) (0.39–1.58)

Table 3 Post-consultation 30-day mortality: factors related to the maternal encounter with health professionals; 1572 sick children seeking consultation at the paediatric outpatient and emergency department of the Simaõ Mendes National Hospital, Bissau, Guinea-Bissau

(0.44–1.77) (0.34–1.34) (0.18–0.93) (0.42–1.75) (0.17–0.95) (0.38–2.02)

Adjusted for assistant-rated severity, hospitalization at the first visit and age of child. àMidwife, laboratory technician, administrative person and cleaner.

87.7% (64/73) of children who died within 24 h of the first consultation and 86.3% (88/102) of children who died within 7 days. Eight of nine children who were not 1872

hospitalized at the first consultation and who died within 7 days of this visit were rated moderately to severely ill by the assistant. Likewise, waiting time to first attention





correlated well with 30-day mortality risk, OR ¼ 0.49 (0.35–0.67), for each group of increasing waiting time. Mothers belonging to Muslim ethnic groups were less likely than other ethnic groups to present a child rated as severely ill by the field assistant [OR ¼ 0.27 (0.12–0.62), age adjusted]. Mothers who previously lost a child were not more likely to bring a child rated severely ill (P ¼ 0.56), nor did they have a greater chance of being hospitalized (P ¼ 0.99). Previous hospitalization did not predict mortality. Children whose parents were sent out to find money and drugs had a twofold increase in mortality risk, but this was explained entirely by the severity rating of the field assistant (Table 2). There was a linear trend in increasing likelihood of being sent out to find money and drugs with increasing severity rating by the assistant (OR ¼ 3.18 per increase in severity group, P-trend < 0.001). This trend remained significant with a P-trend < 0.001 when adjusted for the factors in the final multivariate mortality model. Favouritism and background factors Information on acquaintance and familiarity with health or health-related persons was available for 1384 children, of whom 810 claimed to know such a person and 574 did not. A physician was the health person most commonly known (n ¼ 524) followed by a nurse (n ¼ 364). Acquaintance with both a physician and a nurse was reported by 204 mothers. There was a significant increase in proportions of mothers being acquainted or familiar with a medical doctor with increasing school education (OR ¼ 1.15 per year of school education, P < 0.001). The richest mothers were slightly more likely to know a health person than the poorest mothers [OR ¼ 1.38 (1.02–1.87)]. There was no correlation between being acquainted or familiar with a field assistant from the surveillance project and school education (P ¼ 0.79) or socioeconomic status (P ¼ 0.85). There was no interaction with background factors and acquaintance with a physician. Ethnic groups of Muslim orientation were less likely to be acquainted with a health person [OR ¼ 0.58 (0.38–0.90), adjusted for age of child], compared with other ethnic groups. There was no significant relation between being acquainted or familiar with health personnel and mother registered in the BHP surveillance system during pregnancy (P ¼ 0.49), previous loss of child (P ¼ 0.62), previous hospitalization (P ¼ 0.79) or mother’s age (P ¼ 0.26). Favouritism and 30-day mortality risk Factors related to the clinical care situation between mothers and the staffs in the emergency clinic, in relation


to 30-day mortality, are shown in Table 3. Acquaintance with a medical doctor reduced the 30-day mortality by nearly 45% [OR ¼ 0.55 (0.33–0.92)], adjusted for age of child, assistant-rated severity and hospitalization at the first visit. This effect tended to be stronger among postneonatal children [OR ¼ 0.45 (0.25–0.81)] than neonates [OR ¼ 1.59 (0.34–7.40)], adjusted for assistant-rated severity and hospitalization at the first visit (P ¼ 0.07). Acquaintance with any other health worker or field assistant had no overall impact on survival, but the effect of being acquainted with a nurse was modified by the child’s age: knowing a nurse elevated the mortality risk among infants OR ¼ 2.42 (1.30–4.23), while there was no effect of knowing a nurse among older children OR ¼ 1.00 (0.47–1.97), P < 0.001. Among mothers who knew more than one health person, there was no interaction between the different types of persons known and mortality risk. Mortality varied with consulting physician but was not significantly related to the experience level assessment of colleagues (Table 3). Medical doctors who were associated with low mortality were less likely to belong to the group of doctors who hospitalized children most frequently from the outpatient clinic (58.1%) than doctors belonging to the medium or the highest mortality group (80.7%), P < 0.001. Mortality risk varied to a large extent with clinical team on duty in the outpatient clinic. Unfortunately, team members switched half way through the study period; hence, the variable could not be included in the multivariate analysis. Mortality varied markedly with day of week of consultation and hour of consultation even adjusting for assistant severity rating. Multiple regression mortality model The following variables with a probability of less than 0.10 were included in a multiple regression mortality model: tendency of doctor to hospitalize, sex, socioeconomic index, consultation after 7 pm, consultation elsewhere >48 h, child’s age, mother’s age, ethnic group, mother knows a doctor, day of week, mother multiparous, mother has 9 or more years of schooling, hospitalized at the first visit and assistant severity rating. Waiting time to attention was considered as an intermediate variable associated with the outcome and was excluded from the analysis. In the multivariate analysis, we found acquaintance or familiarity with a physician to be a strong and independent predictor of 30-day mortality reducing mortality by 48% (95% CI: 18–66) (model 1, Table 4). This estimate was not changed if school education and socioeconomic class were forced into the model, although they were significant in the model before knowing that a medical doctor was entered into the 1873




Table 4 Reduced multivariate model for post-consultation 30-day mortality risk; 1572 sick children seeking consultation at the paediatric outpatient and emergency department of the Simaõ Mendes National Hospital, Bissau, Guinea-Bissau Mortality risk Odds ratio (95% CI) Predictor

Model 1

Model 2

Maternal acquaintance or familiarity with a medical doctor Yes 0.55 (0.33–0.94) 0.58 (0.35–1.00) No 1 1 Mother arrived between 7pm and 7 am Yes 1.74 (1.00–3.01) 1.78 (1.02–3.09) No 1 1 Clinical state Severely ill 4.81 (2.00–11.6) 4.87 (2.02–11.8) Moderately ill 1.22 (0.55–2.69) 1.21 (0.54–2.67) Not very ill 1 1 Child consulted elsewhere >48 h,