6
Integrated and General Health Programs

INTRODUCTION

The previous chapters reviewed studies of disease-specific interventions. In practice, few health programs are limited to a single intervention against a single disease. Immunization programs include vaccinations against several diseases (usually tuberculosis, diphtheria, pertussis, tetanus, polio, and measles). Health centers often offer vaccinations, as well as treatment for fever, diarrhea, and tuberculosis, and many provide growth monitoring and health education. Hospitals often include a health center that provides all these services as well as curative services for a wide range of diseases and conditions, including pregnancy.

Integrated services are expected to have an effect on mortality because they generally include some interventions that have demonstrated effects when delivered alone. For example, because most health centers and hospitals provide measles vaccinations, at a minimum these programs have an impact through this one service. However, we would like to know whether the effects of these integrated services are larger than the effects of much simpler (and presumably cheaper) programs based on a few selected interventions.

Integrated health programs, health centers, and hospitals should not be evaluated solely on the basis of their effect on mortality. In the first place, they provide services that reduce morbidity, and they often serve as training centers for doctors and other health personnel. Second, hospitals also have



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 132
6 Integrated and General Health Programs INTRODUCTION The previous chapters reviewed studies of disease-specific interventions. In practice, few health programs are limited to a single intervention against a single disease. Immunization programs include vaccinations against several diseases (usually tuberculosis, diphtheria, pertussis, tetanus, polio, and measles). Health centers often offer vaccinations, as well as treatment for fever, diarrhea, and tuberculosis, and many provide growth monitoring and health education. Hospitals often include a health center that provides all these services as well as curative services for a wide range of diseases and conditions, including pregnancy. Integrated services are expected to have an effect on mortality because they generally include some interventions that have demonstrated effects when delivered alone. For example, because most health centers and hospitals provide measles vaccinations, at a minimum these programs have an impact through this one service. However, we would like to know whether the effects of these integrated services are larger than the effects of much simpler (and presumably cheaper) programs based on a few selected interventions. Integrated health programs, health centers, and hospitals should not be evaluated solely on the basis of their effect on mortality. In the first place, they provide services that reduce morbidity, and they often serve as training centers for doctors and other health personnel. Second, hospitals also have

OCR for page 132
the potential to serve as centers for research into local health problems. Third, the development of efficient health centers and hospitals is a necessary precondition for the continuing improvement of health services in Africa (Mosley, 1988; Newell, 1988). Although many of the individual interventions can be delivered efficiently without extensive investment in health centers and hospitals, that may not be true of the interventions that will be the focus of future efforts to improve health services in Africa (Ewbank and Zimicki, 1988). For example, although vaccinations can be provided through periodic visits of mobile teams, treatment for acute illnesses such as malaria and respiratory infections must be available almost continuously. In addition, as the number of services provided increases, the relative cost-effectiveness of fixed clinics may increase because the fixed costs of the clinic are distributed over a larger number of effective interventions. Therefore, health centers and hospitals should be evaluated in the context of the continuing development of general health services. Evaluation of the efficacy of integrated health programs is much more problematic than evaluation of single interventions. First, it is difficult to assume that the results of a study of one program apply to other programs because there is substantial variation among integrated programs. For example, some vaccination programs include locally important vaccines such as meningococcal vaccine. Similarly, health centers differ in the number of drugs and amount of equipment available, the staffing patterns, and the skill level of personnel. Even if the same services are offered, there might be substantial variation in the mix of services actually delivered. For example, health centers may encourage mothers to return monthly for the series of diphtheria-pertussis-tetanus (DPT) and polio injections, but may find it difficult to have them return for a measles injection after the child is 9 months old. On the other hand, mobile clinics may achieve higher coverage of measles (a single injection) than DPT, which requires three injections each separated by at least one month. A second problem in the evaluation of integrated programs is that it is often difficult to identify and maintain control areas. When national programs begin, there is often an attempt to introduce them in all parts of the country as quickly as possible. For example, the evaluation of the Combatting Childhood Communicable Diseases Project in Zaire began with a program area and a control area. However, after the baseline survey, the program managers decided to combine the two areas into one health zone (Chahnazarian et al., 1993). Even when scientists can define control areas, it is often difficult to limit the use of services to those living in the intervention areas. For example, it is rarely possible to restrict access to a health center to a defined population. Therefore, evaluations of such programs are prone to contamination of the control area.

OCR for page 132
Finally, the more complex the program, the less likely are all of its aspects to be successfully introduced simultaneously. Therefore, there is no exact date for the start of the program. Training of personnel often continues after the start of services, the range of services may increase slowly, and problems in management and staffing may delay the effective operation of the program. In many cases, the program will be phased in over so many years that it is not feasible to separate its effects from long-term trends. Given these difficulties, it is not surprising that few studies have produced reliable estimates of the effect of hospitals, health centers, or integrated programs. CASE STUDIES OF LONG-TERM MORTALITY TRENDS The best evidence for the effect of health programs on mortality in Africa comes from a few studies of long-term trends in mortality in West Africa. Although these studies do not offer the kind of rigorous proof provided by randomized trials or even natural experiments with control areas, they do suggest that the expansion of health services has had an impact on survival rates. Mlomp, Senegal Pison et al. (1993) have examined mortality trends in Mlomp, a rural area in southern Senegal. Maternity histories collected in 1984-1985 were compared with registers from maternity clinics, civil registers, and records from religious missions and health dispensaries. These data provide estimates of infant and child mortality for the period 1930-1984. Although the quality of these estimates cannot match that of a continuous registration system, they provide a reasonable outline of mortality trends. It appears that the proportion of children dying by age 5 was quite stable at about 350-370 per 1,000 from 1930 to 1960. After 1960 mortality began to decline rapidly and fell to only 81 per 1,000 for 1985-1989. Pison et al. provide a convincing case for the conclusion that this rapid decline in mortality and the very low levels today are due largely to the introduction of health services during the 1960s and 1970s. These services include the opening of a dispensary in 1961 and a maternity clinic in 1968. Vaccination programs started in 1971, but were irregular until the late 1970s. By 1980 most children were correctly vaccinated. Ninety-nine percent of the children born in Mlomp in 1988 and still living in the area had received measles, yellow fever, BCG, DPT, and polio vaccines. A growth monitoring program began in 1969 and supplementary foods were provided to mothers and children. In 1975, an antimalarial program began promoting regular doses of chloroquine during the rainy season. Chloroquine is provided free to pregnant women and young children.

OCR for page 132
The effect of these programs can be seen in data on cause of death and in epidemiologic data. In 1963 a survey showed a malaria parasite rate of 50 percent in children. In 1989 the prevalence of malaria parasitemia was only 3 percent among children aged 0-6 years at the end of the rainy season. During 1985-1989 the data on cause of death suggest that the probability of dying of malaria between ages 1 month and 5 years was only 2 per 1,000. The last epidemics of measles occurred in 1972 and 1974. During 1985-1989 there were no measles deaths and only one death from whooping cough. There have been socioeconomic changes in Mlomp, but they were probably not responsible for most of the declines in infant and child mortality rates. Education levels of women have only increased recently, and in 1985-1989 only 26 percent of the births were to mothers who had completed at least one year of schooling. Transportation improved, but not until after the mortality decline was well established. Finally, there was no substantial improvement in the economic situation in the area. The Mlomp example suggests that the provision of basic health services can bring about substantial declines in infant and child mortality. Keneba, The Gambia, 1950-1984 Four villages (Keneba, Manduar, Jali, Kanton-Kunda) located in the West-Kiang district in The Gambia, totaling about 2,000 persons, have been followed by a team of British researchers since 1949 (Billewicz and McGregor, 1982; Lamb et al., 1984). During the first 25 years, there was no evidence of mortality decline in Keneba. If anything there was a small rise in mortality from 1950 until 1970. Mortality was extremely high during this period, with a peak probability of dying before age 5, 5q0, of 488 per 1,000 live births and an infant mortality rate of 220 in 1966-1970. Massive interventions were introduced between 1975 and 1983 in three villages. A physician or qualified nurse-wife was available in the area 24 hours per day. General clinics, as well as prenatal and well-child clinics were held weekly. Women were provided regular gynecological and contraceptive services. To facilitate use of the health services, transportation to and from the clinic was free. Should community members require hospital admission, they were transported to the government hospital. Moreover, traditional midwives were trained to use more hygienic delivery practices. Mortality dropped rapidly. Mortality under age 5 dropped by 75 percent to 110 deaths per 1,000 children in 1982-1983. Infant mortality dropped by 89 percent to only 25 deaths per 1,000 live births. This dramatic decline was achieved by having a physician or a qualified nurse constantly on call. They had a minimal list of essential drugs and means of transportation for referral when necessary.

OCR for page 132
This case suggests that relatively simple and efficient interventions can bring about a major reduction in mortality in a very short period of time, despite the poor socioeconomic conditions and the low level of education that prevailed throughout the period. It is probably the most striking example of what can be achieved with simple and appropriate interventions. Niakhar, Senegal, 1962-1989 A rural area of Senegal (Niakhar) has been under demographic surveillance since 1962. During the first period, 1962-1972, mortality was extremely high with an average rate by age 5 of 500 deaths per 1,000 children (Garenne, 1981). There was evidence of mortality decline for all ages and especially for children under 5 years of age since 1962. The decline was not regular. The first major drop in under-5 mortality occurred between 1970 and 1979. Despite the declining trend, annual fluctuations have been considerable, with rates often doubling from one year to the next, most of them being beyond the 95 percent confidence intervals. An attempt to reconstruct the earlier trends in birth and death rates indicated that mortality has declined since at least 1954 and that fertility began to rise at approximately the same time. A number of interventions were carried out in Niakhar in the late 1970s and through the 1980s. A food supplementation program, which included a growth monitoring component, was carried out between 1972 and 1989. In 1986, a large-scale primary health care program, which included EPI vaccines, oral rehydration therapy, treatment for acute respiratory infections, availability of essential drugs, and medical services, was introduced. Vaccination coverage increased substantially. Measles vaccination coverage increased from 8 percent in 1986 to 80 percent in 1989. The percentage of children receiving the third dose of the DPT vaccine increased from 0 to 56 percent over the same period. Mortality declined dramatically within three years. Child mortality dropped by 52 percent between 1984-1986 and 1989. During the same period, the infant mortality rate dropped by 41 percent. By 1989, the under-5 mortality rate was 159 deaths per 1,000 children and the infant mortality rate (IMR) was only 69 deaths per 1,000 live births. The neonatal mortality rate was less affected by the project, dropping only 26 percent. Most of the mortality decline could be attributed to four causes of death that were the target of the interventions: measles, pertussis, diarrhea, and acute respiratory infections. This case study shows that major annual fluctuations are a normal component of child mortality in Africa, that local trends in mortality are far from linear, and above all, that simple selective public health interventions can have a major impact on cause-specific mortality of children in a very short period of time.

OCR for page 132
HOSPITALS AND HEALTH CENTERS Table 6-1 presents estimates from the Demographic and Health Surveys (DHS) of the proportion of children with diarrhea, fever, or respiratory problems who received treatment at a medical facility. Basic medical services reach a large proportion of the population of many countries. As a result, the potential effect of these programs is substantial. Despite the large investments in hospitals and health centers, there are very few instances in which the mortality impact of these facilities has been estimated. Little or no association between child mortality trends in the late 1970s and the availability of health services was reported for Kenya by Blacker et al. (1987), with the exception of a few districts, as discussed in Chapter 7 of the report of the Working Group on Kenya (1993). The recent DHS in Liberia and Zimbabwe provide an opportunity for a crude test of the effect of health centers. In the Liberia DHS, women reported how long it took them to travel to the nearest clinic. In Zimbabwe, a separate survey was conducted to record the nearest clinic for some of the clusters. Katende (1992) analyzed these data for evidence that proximity to TABLE 6-1 Proportion of Reported Recent Cases of Diarrhea, Fever, and Cough or Difficulty Breathing Among Children 1-59 Months of Age Taken to a Medical Facility, Selected Countries of Sub-Saharan Africa, 1986-1989         Cough or Difficulty Breathinga Country Diarrheaa Feverb Diarrhea and Fever Fever No Fever Botswana 45.9 90.2 69.1 92.4 81.2 Burundi 38.1 49.9 45.7 45.1 34.8 Ghana 43.1 56.4 48.9 51.5 46.3 Kenya 46.8 55.5 49.7 70.7 56.7 Mali 2.8 2.9 3.3 7.3c 3.0c Senegal 19.4 57.6d n.a. n.a. n.a. Togo 25.4 30.8a 28.9 36.2c 23.9c Uganda 14.8 48.3 18.3 57.6 46.8 Zimbabwe 32.9 n.a. 34.6 58.9 53.5 NOTE: n.a. = not available. a Two weeks preceding survey. b Four weeks preceding survey unless otherwise noted. c Rapid breathing with or without fever. d Malaria during the last cold season. SOURCE: Boerma et al. (1991:Tables 3.14, 4.2, and 5.2).

OCR for page 132
a health center is associated with lower infant or child (i.e., ages 1-4 years) mortality. His analysis, which controlled for age, education, socioeconomic status, ethnicity, child immunization status, and water source, did not provide any support for the hypothesis that proximity of health clinics reduce child mortality. The only significant result was for infant mortality in Liberia. Liberian children who live within 30 minutes of a health clinic did appear to have significantly lower infant mortality rates than those who did not. However, this relationship became nonsignificant when several control variables, in particular ethnicity, were included. We do not consider the results of these analyses to be conclusive. First, cross-sectional data are not suitable for rigorous tests of causality. For example, in areas where a new health clinic opened shortly before the survey, some children may have died before the clinic opened. More significantly, if the government placed clinics in areas where mortality was highest, the cross-sectional results would be biased against finding that clinics reduce mortality. Therefore, it is important to compare the timing of changes in health services to the timing of changes in mortality. Second, there may not be sufficient variation in accessibility to a clinic to discern a significant effect. In Zimbabwe, particularly, most women included in the survey were within one hour of a clinic. Although the difference between 30 minutes and 60 minutes of travel time might be significant for minor health problems, it might not affect use of the clinic for life-threatening conditions. Because of these problems, it is necessary to rely on studies that examine changing access to health services rather than those providing information at only one point in time. PRIMARY HEALTH CARE PROGRAMS Village Health Workers Greenwood et al. (1990) have measured the effect of the primary health care (PHC) program that began in The Gambia in 1981. The program relied on the work of village health workers (VHWs) and traditional birth attendants (TBAs). The responsibilities of the VHWs included treatment of malaria, diarrhea, and acute respiratory infections, and promotion of improved nutrition and immunization. The VHWs received an initial supply of basic drugs including aspirin, chloroquine, oral penicillin, an antacid, and an anthelmintic. They were to buy replacements with funds obtained through the sale of drugs to their patients. In 1983 the PHC program began in most of the villages in Farafenni district having a population of at least 400. The ''non-PHC" villages served as controls. All villages had access to a health center in Farafenni town. Data on program activities suggest it is unlikely that the VHWs had a

OCR for page 132
large effect on mortality. For example, the VHWs made an average of only one or two visits per child per year. This number was probably too small to have brought about a large mortality decline, especially because there apparently was a perception that the VHWs were not responsible for curative services. Vaccination coverage was already high before the program began (e.g., measles coverage rates of 92 percent in the PHC and 84 percent in the non-PHC villages). Therefore, there was little room for the VHWs to increase vaccination coverage. In addition, the VHWs had only a small impact on the proportion of deceased children who were seen by a doctor before their death. This proportion increased from 48 to 57 percent. A comparison of the infant and child mortality rates (CMRs) in The Gambia for the year before the start of the program with the rates for the three years after the start show modest, nonsignificant changes. A comparison of the baseline year with the third year after the start of the program does show a significant decline in the IMR. However, the decline in the CMR is still not significant. In addition, there were similar declines in mortality in the non-PHC villages. The results of the study of PHC in The Gambia are disappointing. However, the sample sizes involved are such that it would be surprising if the differences between PHC and non-PHC villages were significant. If mortality in the PHC villages was actually 20 percent lower than mortality in the non-PHC villages, there would be only a 26 percent chance of finding a difference in infant mortality that was significant at the 5 percent level. There would be only a 30 percent chance of finding such a difference in under-5 mortality. A 20 percent reduction in mortality would be a large reduction, given the differences in level of coverage described above. Therefore, it is not surprising that results fail to show significant differences in mortality. In addition to the small sample size, it appears that children in non-PHC areas received treatment from VHWs from PHC villages, which would bias the result toward finding no difference in mortality between PHC and non-PHC villages. Pahou Primary Health Care Project The Pahou PHC project (Velema et al., 1991) was a field test of an approach to primary health care in Benin. The project covered 16 coastal villages with a population of 13,000 about 30 kilometers from the national capital, Cotonou. The program involved 17 VHWs and vaccination teams that visited each village. The VHWs were trained to visit families with young children and to provide malaria chemoprophylaxis (pyrimethamine) and ORT packets when needed. They were also trained to provide curative services including chloroquine for malaria, sulfadimidine for respiratory

OCR for page 132
infections, mebendazole for parasitic infestations, ORT for diarrhea, and topical aureomycin for eye infections. They also referred patients to the communal health center. The quality of data collection improved over time, and it is not possible to calculate reliable measures of program impact. In addition, the project did not include a control area. However, Velema et al. (1991) carried out a case-control study to estimate the efficacy of various interventions. The study involved 74 children who died between ages 4 months and 3 years. The deceased children were matched with 230 controls by month of birth, sex, and village of residence. Any interventions received by the controls after the death of the case child were excluded from the analysis. The intervention associated with the largest reduction of mortality was measles vaccination before 12 months of age. The relative risk for children vaccinated early was 0.36 (95 percent confidence interval (C.I.) 0.16-0.81) compared to unvaccinated children. Measles vaccination after 12 months of age was not associated with any reduction in risk. This curious finding suggests there may be other factors associated with the age at vaccination that contribute to these results. However, the authors state that their results did not change when they included all study variables (among them, apparently, measures of socioeconomic status) in a single logistic regression. The other EPI vaccinations—DPT, polio, and bacille Calmette-Guérin (BCG)—were not associated with any reduction in mortality. However, the length of the study was so short (34 months) that diphtheria, pertussis, and polio were not prevalent during the study period (Velema et al., 1991). A longer period of observation might have found a greater effect of these vaccinations. Children who received only a single DPT immunization (DPTI) had an elevated risk of death. However, this probably resulted from DPTI shots given to children who came to the health center for curative services. These sick children were at increased risk of death and were less likely to receive a second DPT. Estimates of the efficacy of VHW visits applied only to those villages where there was a VHW, thus excluding the two largest villages, which were close to the communal health centers. More than 70 percent of cases and controls were seen by a VHW. Although cases were less likely to have seen a VHW, this difference was not significant. Similarly, the risk of death decreased with the number of contacts with VHWs, but this difference was not significant. However, significantly more controls than cases had been seen by a VHW in the six months preceding the death of the case. Those who had seen a VHW in the past six months had a relative risk of death of 0.33 (95 percent C.I. 0.16-0.69) compared to those who never had contact with a VHW. The number of visits in the most recent six months was also significant. Unfortunately, the authors do not provide a simple comparison of those

OCR for page 132
TABLE 6-2 Relative Risk of Death Associated with Timing of VHW Visits in Months Preceding Death, Pahou Primary Health Care Project, Benin, 1986-1987   Relative Risk 95% C.I. No visit ≤ 6 months 1.00   Visits only ≤6 months 0.39 0.16-0.97 ≤6 and 7-12 months 0.45 0.16-1.28 ≤6, 7-12, and > 12 months 0.30 0.09-0.97 χ2, 3 df 8.71     SOURCE: Velema et al. (1991). By permission of Oxford University Press. who had a visit in the past six months with those who did not. In one test, they examined the relationship between mortality and the number of visits. In another test, they compared those who did not have a visit in the most recent six months with those who had a recent visit, but had not had a visit more than six months ago. The results of the latter analysis are summarized in Table 6-2. The authors concluded that there is evidence of "a linear trend of increasing protection with increasing regularity of contact with the VHW (χ2 for linear trend: 7.83 on 1 df [degree of freedom])" (Velema et al., 1991:477). However, a reanalysis of the data suggests that the relevant difference is between those who had a recent contact with the VHW (within six months of the age at death of the case) and those who did not. The Pahou study is an innovative attempt to demonstrate program impact using a case-control design. This approach has the additional advantage of providing some information on the impact of individual interventions. However, the results must be considered tentative because of the likelihood that there are unmeasured differences between families that might affect both mortality and contact with the VHWs. Social class and attitudes toward modern health services are two important sources of variation in both mortality and use of modern services. These factors might be important sources of confounding. CHILD SURVIVAL AND EXPANDED PROGRAMS ON IMMUNIZATION Combatting Childhood Communicable Diseases (CCCD) is a program funded by the U.S. Agency for International Development to provide support

OCR for page 132
to child survival programs in Africa. The program provides assistance for increasing vaccination coverage, home-based use of oral rehydration therapy for diarrhea, and presumptive treatment of fevers with chloroquine in malarious areas. Several of the studies quoted in this report were carried out by researchers and program staff associated with CCCD programs (e.g., Cutts, 1988; Taylor et al., 1988; Deming, 1989; Cutts et al., 1990a-b, 1991). In addition, CCCD funded two studies of the effect of child survival activities in areas of two CCCD countries. One was a study of the effect of the national CCCD program on the health services and mortality in Bomi and Grand Cape Mount counties in Liberia. The other examined the effect of the CCCD program in Zaire in the Kingandu area. These studies have the advantage of reporting on the effectiveness of two national programs rather than special demonstration projects. Evaluation of CCCD in Liberia Between 1984 and 1988, the CCCD program in Liberia increased coverage with three shots of DPT from 1 to 15 percent and coverage of measles vaccine from 13 to 33 percent among children aged 12 to 23 months. The proportion of pregnancies protected by two injections of tetanus toxoid increased from very low levels to more than 30 percent (Foster et al., 1993). Becker et al. (1993) reported on the results of two surveys carried out in 1984 and 1988, which recorded mortality rates before and after the start of the program. A comparison of the two years preceding the start of the program with the subsequent two years shows that infant mortality declined by an estimated 24.5 percent (from 240 to 181 death per 1,000 live births). Mortality at ages 1-4 years declined by 28 percent (from 46 to 33 deaths per 1,000). Both changes were significant at the 5 percent level. The study did not include a control area since the program was introduced nationally. However, it is unlikely that declines of this magnitude would have occurred in the absence of the program. Although it appears that the program reduced mortality overall, the studies did not provide convincing evidence on which components of the program were responsible for the declines. Estimates of the cause of death suggested that mortality due to neonatal tetanus declined by more than 50 percent. Coverage with tetanus toxoid most likely did not increase this much so the decline in tetanus mortality was probably not quite as large. The other cause of death that showed a decline was "fever." However, reported used of antimalarials did not change over the period (Foster et al., 1993). There was no decline in the reporting of measles deaths, nor was there any relationship between the increase in measles coverage by survey cluster and the change in mortality at ages 1-4 years. The failure of the CCCD study in Liberia to demonstrate declines in

OCR for page 132
causes of death that were targeted by the program is disappointing, but perhaps not surprising. The verbal autopsies employed in the study were taken from questionnaires used in the Philippines. It is not clear to what extent they were valid in Liberia. In addition, the Liberia study used slightly different criteria for diagnosis. For example, in the Philippines, the criteria for diagnosis of nonfatal measles were "age greater than 120 days, rash and fever for more than three days." These criteria had a sensitivity of 98 percent and a specificity of 90 percent (Kalter et al., 1990). Becker et al. (1993) listed their criteria for diagnosis of measles as "presence of skin rash and child over 2 months of age." By dropping fever and not limiting the rash to those lasting at least three days, the specificity of the criteria was probably lower than it was in the Philippines. In addition, dropping the age limit from 4 to 2 months probably led to an exaggeration of measles deaths at ages 1-5 months. If the sensitivity and specificity of the criteria used in the CCCD study in Liberia were the same as those estimated by Kalter in the Philippines, only about 59 percent of the reported measles deaths in the 1984 were in fact measles.1 This proportion is termed the "positive predictive value" of the criteria. If the specificity was only 85 percent (because of the less stringent criteria used in Liberia), then only 34 percent of the reported measles deaths were actually due to this cause. With such low positive predictive values, comparisons of reported measles mortality rates before and after the start of the CCCD program do not provide a powerful test of the hypothesis that the program reduced measles mortality. Because measles immunization coverage increased from 13 to only 33 percent, we might have expected true measles deaths to decline by roughly 20 percent. With a positive predictive value of 59 percent, the expected decline in reported measles deaths would be only 59 percent of 20 percent, or approximately 12 percent, or a decline from about 18 to 16 deaths per 1,000. There are similar problems with their other reports of causes of death, such as acute respiratory infections. For example, Kalter et al. (1990) estimated that the sensitivity and specificity of "cough for at least 4 days and dyspnoea for at least 1 day" are 59 and 77 percent, respectively. Becker et al. (1993) used the criteria of "cough and trouble breathing for more than 2 days." The positive predictive value for these criteria is probably not high enough to be useful for studying changes in the mortality rate due to respiratory infections. 1   The true proportion of all deaths due to measles can be estimated as (R + - 1)/( + - 1) where R is the reported proportion due to measles, is the sensitivity, and is the specificity.

OCR for page 132
Evaluation of CCCD in Zaire The CCCD program in Zaire succeeded in increasing vaccination coverage in the Kingandu area. Measles immunization coverage increased from 22 to 74 percent, and coverage with three doses of DPT increased from 15 to 62 percent (Vernon et al., 1993). Reported use of oral rehydration increased from less than 6 percent to greater than 55 percent per episode. The proportion of pregnancies protected by two injections of tetanus toxoid increased from 14 to 43 percent. There was an evaluation of the effect of the CCCD program in Zaire in Kingandu (Chahnazarian et al., 1993). The mortality rate among children aged 0 to 5 declined from 41 per 1,000 during the five years preceding the start of the program to 33 during the next five years. This decline was concentrated at ages 1 to 4 years, where mortality declined by 33 percent (95 percent C.I. 22-45 percent). A regression analysis suggested that this decline is more consistent with a program effect than with a steady downward trend, although it is not possible to disentangle the two possibilities statistically. One approach to determining whether the decline in mortality was due to the program is to examine changes in the reported number of measles cases (Chahnazarian et al., 1993). There was a sharp drop in the annual number of measles cases reported at the local hospital following the start of the program and the increase in vaccination coverage. During 1978-1984, there was an average of 108 measles cases at the hospital each year. After the start of the CCCD program (1985-1989), there was an average of 36 cases per year. Most of these cases occurred during an outbreak in the hospital in 1988. A regression of the mortality rate at ages 6 to 35 months on the reported measles cases for the years 1978-1987 and 1989 shows that variations in the number of reported measles cases explained 67 percent of the variance in mortality. Therefore, the decline in the average annual number of measles cases in the hospital between the preprogram period (1978-1984) and the postprogram period (1985-1989) was associated with an 18 percent decrease in mortality at 6 to 35 months. This estimate probably included some effect from the other program interventions. However, it is likely that the largest share of the effect is attributable to measles immunization. SUMMARY It is difficult to believe that the increasing availability of modern health services provided at hospitals, at health centers, and through integrated health programs has not played some role in the long-term decline in infant and child mortality in sub-Saharan Africa. However, there is very little

OCR for page 132
evidence to help us determine whether their contributions have been trivial or substantial. These programs and services are difficult to evaluate because they are based on a set of services rather than a single intervention and are usually introduced slowly over a period of time. Case histories from Senegal provide some evidence that the declines in mortality were associated with increases in health services. The data from Keneba in The Gambia are more convincing because the health programs were introduced over a period of a few years and were associated with a very rapid decline in mortality. The evaluations of the CCCD programs are valuable because they are the most recent studies of the effects of national integrated health programs in Africa. Both studies suffer from lack of control areas. In Liberia there were no opportunities for controls, and in Zaire the control area was administratively incorporated into the program area after the start of the study. However, both studies suggested large declines in mortality that were temporally associated with the start of the program. There is little evidence about the effect of village health workers on mortality. We have not attempted to review all of the literature on the effect of VHW schemes on vaccination coverage, the provision of services by VHWs, and the frequency and timeliness of referrals to clinic. However, the studies reviewed here as well as others (e.g., Nougtara et al., 1989) suggest that VHWs do not necessarily increase utilization of health services. Many more studies of the effect of primary health care on mortality are needed. Because there is probably great variability in the performance of general health programs, we need more small-scale studies (such as that in The Gambia). Planning small studies will improve the feasibility of completing them. Increasing the number of studies will reveal whether there are any generalizations that can be made about this type of program. Whenever possible, these studies should examine the longer-term effect of programs because it may take several years for them to achieve a level of operation that shows an effect.