Annals of Tropical Paediatrics (1999) 19, 135± 142
Paediatric hospital admissions at a South African urban regional hospital: the impact of HIV, 1992± 1997 KAREN J. ZWI, JOHN M. PETTIFOR & NEIL SODERLUND* Department of Paediatrics, Chris Hani Baragwanath Hospital and the University of the Witwatersrand, and *Centre for Health Policy, University of the Witwatersrand, South Africa (Accepted 26 November 1998)
Summary Rates of infection by the human immunode® ciency virus (HIV) have been increasing rapidly in South Africa over the last decade. This study documents the changes over time in prevalence of HIV infection amongst hospitalized children, and its effects on the pro® le of disease and in-hospital mortality over the period 1992± 1997. Admissions to the paediatric medical wards between January 1992 and April 1997 were obtained from the routine computerized database held in the Department of Paediatrics at Chris Hani Baragwanath Hospital. HIV tests were performed on clinical indications only. Over the study period there were 22,633 admissions involving 19,918 children. Total annual admissions increased by 23.6% between 1992 and 1996. Prevalence of HIV infection increased from 2.9% in 1992 to 20% in 1997. HIV-infected children had a younger age distribution, longer median length of stay and more readmissions (p , 0.001) compared with HIV-negative and untested children. HIV-infected children accounted for the increased number of admissions for pneumonia, gastro-enteritis, malnutrition and tuberculosis, and the rise in in-hospital mortality by 42% from 4.3% in 1992 to 6.1% in 1997. Paediatric HIV infection has changed the pro® le of paediatric admission diagnoses and increased in-hospital mortality in the relatively short time between 1992 and 1997. Over the same period, HIV-negative children showed declining rates of malnutrition, vaccine-preventable diseases and admission to the intensive care unit.
mission are not available in the public health service except in a few centres where clinical trials are being conducted. Chris Hani Baragwanath Hospital is the largest in the world, with more than 3,000 beds, 7,000 members of staff and over 100,000 admissions per year.3 It is one of 40 provincial hospitals ® nanced and run by Gauteng, South Africa’ s most densely populated province. It serves Soweto, south-west of central Johannesburg, an urban working class community of approximately one million people. It acts both as a local community hospital and as the tertiary referral hospital for southern Gauteng. This study was performed to document the impact of paediatric HIV on hospital admis-
Introduction In South Africa there are an estimated 2.5 million people currently infected with HIV. In Gauteng, the province in which is located the regional urban hospital under study, Chris Hani Baragwanath, seroprevalence among antenatal clinic attendees was 15% in 1996.1 Vertical transmission in South Africa is estimated to be between 26% and 42%.2 Antiretroviral medication and elective caesarean section to reduce mother-to-child transReprint requests to: Dr Karen Zwi, Division of Community Paediatrics, Department of Paediatrics, Chris Hani Baragwanath Hospital, PO Bertsham 2103, South Africa. Fax: 1 11 938 8162; e-mail:
[email protected] 0272-4936/99/020135-08 $9.00 Carfax Publishing
Ó
1999 The Liverpool School of Tropical Medicine
136
K. J. Zwi et al.
sions at Chris Hani Baragwanath Hospital between 1992 and 1997, a period during which paediatric HIV infection in the region increased considerably. This is one of only a few hospitals in South Africa with computerized data on paediatric admissions dating back to 1992, providing the unique opportunity to examine trends in hospital admissions over this period. The study objectives were to ascertain HIV seroprevalence in children hospitalized during the study period, to determine the pro® le of hospital admissions, to establish whether any changes over time had occurred, and to determine the impact of HIV disease on child deaths in hospital. In addition, the study aimed to assess whether health trends in children not infected by HIV had improved over time. The study focuses on the paediatric ward admissions at Chris Hani Baragwanath Hospital and excludes the short stay facility and gastro-enteritis unit to which are admitted children with uncomplicated gastro-enteritis or mild illness likely to require a short stay in hospital. Methods The study period was from January 1992 to the end of April 1997. Patient information was obtained from the routine computerized database in the Department of Paediatrics which is generated from entering information from the summary completed by the doctor when a child is discharged, transferred out of the general wards or dies. The discharge summary contains patient demographic characteristics, nutritional status, dates of admission, discharge or death, transfer to the Intensive Care Unit (ICU) and primary and secondary diagnoses coded using ICD 9 codes. HIV status was recorded on the discharge summary when a child was found to be infected with HIV. Children were tested only when clinically indicated and routine screening was not performed on admission, apart from that done as part of a study in one of the four wards for the last 6 months of 1996 (T. Meyers, unpublished data). Informed consent was required
from carers prior to testing. Patient con® dentiality was maintained for the purposes of the study by removing patient names from the researcher’ s dataset. Unique person identi® ers were recorded on the database so that different admissions for the same individual could be linked. The data were analysed using the Statistical Package for the Social Services software (SPSS).4 Comparisons between groups and distributions were made using appropriate statistical methods: c 2 tests, one-way analysis of variance (t-tests) and the generalized linear ANOVA model. Trends over time were analysed using the c 2 test for trend. The data were not available beyond April 1997 at the time of data analysis. Where absolute numbers are presented, complete years 1992 through 1996 are used; where proportions are presented, the 1st 4 months of 1997 have been included. De® nitions were as follows. (i) An HIVpositive child was de® ned as a child of more than 15 months of age who had had two positive HIV ELISA tests or, if less than 15 months, clinical signs suggestive of HIV infection (using the Centers for Disease Control and Prevention’ s 1994 revised classi® cation system)5 together with two positive HIV ELISA tests. (ii) Children were de® ned as having tuberculosis if this was recorded as one of their diagnoses on the discharge summary sheet. Many of these children would have been diagnosed on clinical or radiological grounds, without laboratory or histological evidence of tuberculosis. (iii) Children were classi® ed as malnourished (according to Wellcome criteria)6 if the discharge summary categorized them as having marasmus, kwashiorkor, marasmic-kwashiorkor or as being underweight. Results During the study period there were 22,633 admissions involving 19,918 children. The annual number of admissions increased by 23.6% (from 3,800 to 4,694) between 1992 and 1996, entirely attributable to greater num-
The impact of HIV on paediatric admissions
137
FIG . 1. No. of admissions per year for HIV-negative and untested children, HIV-positive children, and for all children.
bers of HIV-infected children being admitted (Fig. 1). The proportion of children infected with HIV rose almost sevenfold during the study period (from 2.9% of admissions in 1992 to 20% in 1997). Over the study years, 7,604 children (38.2% of all children) were tested. The age distribution of admissions was skewed towards younger children; 49% of all children were under 1 year of age. This was most striking in HIV-infected children, 67% of whom were under 1 year. Median age on admission was 13 months for HIV-negative and untested children and 7 months for infected children. Pneumonia was the most common admission diagnosis in all children and comprised 36.9% of admissions over the study period. The proportion of HIV-positive children admitted with pneumonia was greater than the proportion of negative and untested children, 57.9% versus 34.3% children (c 2 483; p , 0.001). Over the years, the absolute number of admissions for pneumonia rose steadily from 1307 in 1992 to 1902 in 1996, attributable to pneumonia admissions in HIVpositive children (Fig. 2). Gastro-enteritis comprised 21.6% of admissions overall, with HIV-positive children having a slightly higher probability of this diagnosis, 29.4% versus 20.7% in negative and untested children (c 2 90; p , 0.001). Ab-
solute numbers with gastro-enteritis also increased over the 5-year period, but not as strikingly as for pneumonia. Pneumonia and gastro-enteritis declined steadily with age, children under the age of 1 year comprising the majority of admissions. HIV-positive children were far more likely to have a combined diagnosis of pneumonia and gastro-enteritis, which comprised 9.7% of their admission diagnoses, compared with 3.7% of HIV-negative and untested children (c 2 192; p , 0.001). Proportions of children admitted with a clinical diagnosis of tuberculosis increased over the study years from 3.24% in 1992 to 8.26% in 1997. Eleven per cent of HIV-positive children were diagnosed as having tuberculosis compared with 5% of negative and untested children (c 2 147; p , 0.001). Forty-® ve per cent of admitted HIV-positive children were malnourished compared with 22.2% of negative and untested children (c 2 583; p , 0.001; OR 2.8: 95% CI 2.6± 3.1). The proportion with malnutrition declined over time in the negative and untested group, from 26.6% in 1992 to 19± 20% in the later study years (c 2 trend 216; p , 0.001) (Fig. 3). HIV-infected children were more likely to be underweight (19.1%) and marasmic (13.9%), whereas HIV-negative children tended to be underweight (15.8%) or to have kwashiorkor (10.6%). Admissions to the Intensive Care Unit de-
138
K. J. Zwi et al.
FIG . 2. The absolute nos of admissions with tuberculosis, pneumonia and gastro-enteritis by study year for HIV-negative and untested admissions, and HIV-positive admissions.
FIG . 3. The percentage of admissions with malnutrition by study year and for all years combined for HIV-positive, HIV-negative and untested admissions.
clined substantially over the study years, from 3.7% of admissions in 1992 to 1.2% in 1997 (c 2 trend 43; p , 0.001) despite no change in bed numbers or admission policy. Only 23 children of the total 509 ICU admissions were infected with HIV. This re¯ ects Chris Hani Baragwanath ICU policy, which excludes children known to be infected with HIV. The proportion of all admissions who died in hospital increased by 42% over the study period, from 4.3% in 1992 to 6.1% in 1997
(Fig. 4). This is due to the higher overall mortality in HIV-infected children (13.2% versus 5.1% in uninfected and 3.1% in untested children) (p , 0.001) and increasing proportions of infected children. In contrast with this, mortality declined over time in HIV-uninfected children (c 2 trend 3.3; p 5 0.06). The proportion of all HIV-positive children who died increased strikingly from 6.7% in 1992 to 54.3% in 1997 (p , 0.001). The most common causes of death were pneumonia
The impact of HIV on paediatric admissions
139
FIG . 4. The percentage of children who died during hospitalization in each study year for HIV-untested and negative children, HIV-positive children and all children combined.
(24.6%), septicaemia (12,2%), gastro-enteritis (9.5%), nutritional causes (6.1%) and meningitis (4.9%). In HIV-negative and untested children, infectious and nutritional causes of death declined in absolute numbers between 1992 and 1996, but these trends were not signi® cant at the 5% level (p 5 0.4; p 5 0.3). Length of stay in hospital fell over the study period for all children (p , 0.001). HIV-infected children stayed substantially longer than untested and negative children in the early study years, but the difference remained of statistical signi® cance throughout all years (p , 0.001) (Fig. 5). On average, HIV-infected children stayed 1.5 days longer than uninfected and untested children. HIV-infected children were more than twice as likely to have multiple admissions compared with uninfected children, 26.2% being readmitted during the study period, on average twice per year. In children who were not infected with HIV, there was a decline over the study period in the proportion of admissions for immunization-preventable diseases as a group (diphtheria, pertussis, tetanus, poliomyelitis and
measles) (p 5 0.087), measles alone (p , 0.05) and rheumatic heart disease (acute and chronic) (p , 0.001). In contrast with these trends, asthma as a proportion of admissions increased from 3.2% in 1992 to 4.3% in 1996 and 4.1% in 1997 (p 5 0.04) (Table I).
Discussion Total paediatric medical admissions to this urban regional hospital have increased by over 20% in the 5-year study period, entirely attributable to the increasing number of HIV-infected children. In contrast, the number of admissions of non-HIV-infected children has remained stable over this time. As a direct result of the increased numbers of HIV-infected admissions, there has been a shift in the age distribution towards younger children, increased admissions for pneumonia, gastro-enteritis, malnutrition and tuberculosis, more children requiring multiple admission and higher in-hospital mortality rates. At the same time, there has been progressive shortening of the median length of stay and this may re¯ ect
140
K. J. Zwi et al.
FIG . 5. The median duration of hospital stay in each study year for HIV-positive admissions, and HIV-negative and untested admissions.
the need for earlier discharge to make way for the increased patient load. The prevalence of HIV infection among hospitalized children in this study has increased almost sevenfold, from just under 3% in 1992 to 20% in 1997. These ® gures represent a low estimate of HIV seroprevalence as the children studied were tested only on clinical indications and were not routinely screened. The increase in HIV seroprevalence in hospitalized children parallels the increase in HIV infection among women of child-bearing age which has been documented throughout South Africa.7
In line with trends in adults and children in many countries,8± 11 admissions for tuberculosis have more than doubled during the study period in all children, but the increase is more marked in the HIV-infected group. This may be due to real increases in the prevalence of tuberculosis in the adult population, the infecting source of children, or may be as a result of heightened awareness of tuberculosis as a possible diagnosis and more children than previously being given a presumptive diagnosis. Almost half of all HIV-infected children in this study were malnourished. This represents
TABLE I. Selected diagnoses as a proportion (%) of admissions in HIV-negative and untested children (® gures for 1992 and 1997 only)
Diagnosis Measles alone Immunization-preventable diseases Rheumatic heart disease Asthma Congenital abnormalities Kerosene ingestion Other forms of poisoning
1992
1997
2.3 2.6 1.7 3.2 3.5 0.8 2.1
0.5 1.1 0.6 4.1 3.2 0.7 2.4
p-value for trend 1992± 1997 0.016 0.087 , 0.001 0.04 0.24 0.72 0.79
The impact of HIV on paediatric admissions a higher proportion of malnutrition amongst HIV-infected children than seen in other studies in which 17± 24% were malnourished,12± 14 and may re¯ ect that children with more advanced disease are referred to Chris Hani Baragwanath Hospital. Invasive nutrition supplementation procedures such as gastrostomy feeding tubes are not offered to these children. Not surprisingly, over the study period HIVinfected children did not show the decline in malnutrition rates seen in HIV-negative and untested children. HIV-positive children were twice as likely to be readmitted and also stayed longer in hospital (by 1.5 days) compared with HIV-negative and untested children. Their age distribution was shifted towards a younger age at which more intensive nursing care is needed. Bed occupancy rates were over 100% for a substantially greater proportion of 1997 compared with 1994 (J. Pettifor, unpublished data). The numbers of doctors and nurses did not increase during this period to cope with the additional load which implies that medical staff:patient ratios fell and staff workload increased. One might expect that higher bed occupancy rates, reduced staff:patient ratios and increased workload would have some impact on quality of care of hospitalized children. However, adverse events do not appear to be increasing in frequency for HIV-negative and untested children in that mortality rates, ICU admission rates and readmission rates have not increased. More subtle aspects of quality of care are dif® cult to measure and are impossible to ascertain from these data. The pressure on beds may lead doctors to discharge patients earlier than they otherwise would and this may impact on time available for rehabilitation, patient education and counselling. Indeed, the reduced length of hospital stay over the study period may re¯ ect that the pressure on beds is affecting the timing of discharge. The ® nancial implications of the increases in HIV-related admissions relate not only to increased bed occupancy, longer hospital stay and the need for more staff but also to the cost of medication offered to these children. Treat-
141
ment of HIV-infected children with antiretroviral drugs to prolong and improve quality of life is prohibitively expensive in South Africa. However, even if these are not offered, many children infected with HIV require expensive and prolonged courses of drugs. With regard to routine treatment, it is imperative that protocols be affordable and cost-effective to prevent massive escalation in costs. In this study, HIV infection was responsible for a 42% increase in overall in-hospital paediatric mortality and has reversed a trend towards lower rates of in-hospital mortality. Elsewhere in the world, estimates of child mortality, which was projected to decrease in the early 1980s due to improved immunization and primary health care, have been revised upwards as a result of HIV infection.9 Whereas the United Nations’ estimate (without HIV/AIDS) for the under-5 mortality rate in this region was 132/1,000 by the year 2,000, HIV could cause the under-5 mortality rate to rise to between 159 and 189/1,000, an increase of between 20% and 43%.15 The impact of HIV disease on child mortality in hospital is alarming. In 1992, about 7% of deaths were HIV-related, whereas by 1997 HIV-infected children accounted for over half of all ward deaths. At Chris Hani Baragwanath Hospital, 75% of all deaths were HIVrelated in the 1st months of 1998 (U.K. Kala, personal communication). These are trends that are likely to continue well into the future. In contrast with HIV-infected children, the total number of admissions per year of uninfected children has remained stable over the study period. Uninfected children have shown substantial declines in malnutrition rates, rates of vaccine-preventable diseases especially measles, and rheumatic heart disease, a ® nding consistent with studies in other developing countries.16 In uninfected children, there has been a slight decline in in-hospitality mortality and fewer admissions to the Intensive Care Unit. Together with the increase in admissions for asthma and other chronic diseases, this may herald the beginning of a shift towards a more developed world pro® le of disease.
142
K. J. Zwi et al.
The dataset used included only children admitted to the paediatric medical wards and excluded less ill children who could be managed in short stay units. The latter group is less likely to have HIV infection and therefore their exclusion may in¯ ate the HIV infection rates. However, the HIV rates are likely to be an underestimate of the true rates of HIV infection in hospitalized children as they were tested only on clinical indications of HIV infection. In a study of children admitted to Chris Hani Baragwanath in 1996 in which 92.7% of children (those with informed consent) admitted to one of the paediatric medical wards were screened for HIV, 28% of children were found to be infected (T. Meyers, unpublished data). This represents a higher detection rate (by about 8%) than that estimated using the computerized data for that year. Since patients come from a wide area and accurate population denominators are not available, it is not possible to calculate prevalence or mortality rates on a population basis. Considering the burden on the health service due to paediatric HIV disease, it is imperative that measures be introduced to curtail the acquisition of new HIV infections and to implement interventions used elsewhere in the world to reduce mother-to-child transmission. Acknowledgments We thank Nancy Sehlako and Priscia Pillay for data entry and Professor Heather CreweBrown and staff in the South African Institute for Medical Research laboratory at Chris Hani Baragwanath Hospital for providing access to HIV results (in accordance with the University of the Witwatersrand Ethics Committee stipulations).
2
3 4 5
6 7
8
9 10 11
12
13
14
15
References 1 Department of Health and Population Development, South Africa. Seventh national HIV survey of women attending antenatal clinics of the public
16
health services in the Republic of South Africa, October/November, 1996. Epidemiol Comments 1996; 23:4± 16. Bobat R, Coovadia H, Coutsoudis A, Moodley D. Determinants of mother-to-child transmission of human immunode® ciency virus type 1 infection in a cohort from Durban, South Africa. Pediatr Infect Dis J 1996; 15:604± 10. Baragwanath Yearbook, 1996± 1997. Johannesburg, South Africa: Electronic Printing Co., 1997. SPSS for Windows, Version 7.5. Chicago, IL: SPSS Inc., 1996. Centers for Disease Control and Prevention. 1994 revised classi® cation system for human immunode® ciency virus infection in children less than 13 years of age. Morbid Mortal Weekly Rep 1994; 43:1± 19. Wellcome classi® cation of infantile nutrition (editorial). Lancet 1970; 2:302± 3. Department of Health and Population Development, South Africa. Seventh national HIV survey of women attending antenatal clinics of the public health services in the Republic of South Africa, October/November, 1996. Epidemiol Comments 1996; 23:4± 16. Mhalu FS, Lyamuya E. Human immunode® ciency virus infection and AIDS in East Africa: challenges and possibilities for prevention and control. East Afr Med J 1996; 73:13± 19. Coulter JBS. HIV infection in African children. Ann Trop Paediatr 1993; 13:205± 15. Bye RM. HIV infection in children. Clin Chest Med 1996; 17:787± 96. Wilkinson D, Davies GR. The increasing burden of tuberculosis in rural South AfricaÐ impact of the HIV epidemic. S Afr Med J 1997; 87:447± 50. Chintu C, Luo C, Bhat G, et al. Impact of the human immunode® ciency virus type-1 on common paediatric illnesses in Zambia. J Trop Pediatr 1995; 41:348± 53. Vetter KM, Djomand G, Zadi F, et al. Clinical spectrum of human immunode® ciency virus disease in children in a West African city. Pediatr Infect Dis J 1996; 15:438± 42. Bamgboye EA, Familusi JB. Mortality pattern at a children’ s emergency ward, University College Hospital, Ibadan, Nigeria. Afr J Med Med Sci 1990; 19:127± 32. Preble EA. Impact of HIV/AIDS on African countries. Soc Sci Med 1990; 31:671± 80. Petit PLC, van Ginneken JK. Analysis of hospital records in four African countries, 1975± 1990, with emphasis on infectious diseases. J Trop Med Hyg 1995; 98:217± 27.