published in the Asia-Pacific Population Research Reports, Number 9, August 1996, East-West Center, Program on PopulationCONTENTS: * Abstract * Introduction * Data and Methods * Findings * Discussion * References
Using data on infants and children from the 1993 National Demographic Survey of the Philippines, this report investigates the conditions under which young children remain healthy or become ill with acute respiratory infection (ARI) or diarrhea and either receive or do not receive treatment. It focuses on three types of outcome: the current morbidity status for ARI and diarrheal disease of children under five years of age, the type of health care services used (if any) by parents on an ill child, and the mother's knowledge about and use of oral rehydration therapy (ORT). The objective is to identify the major social and economic determinants of these outcomes with a view toward improving health-care programs and ultimately children's survival prospects. The results are presented under the thematic headings of social and economic development, "culture," family and gender relations, parental underinvestment in children and situational factors. The report concludes with a discussion of the policy implications in the study's findings.
If the practice of mortality analysis can be described as relatively undeveloped in the Philippines (e.g. Flieger, Abenoja and Lim, 1981, p.2), then the literature on morbidity patterns among Filipino infants and children must be judged as virtually nonexistent. This is regrettable because the vast majority of childhood deaths must first pass through one or more intermediary disease states. As Mosley and Chen (1984, p.41) have argued, "child mortality should be studied more as a chronic disease process with multifactoral origins than as an acute, single-case phenomenon."
An unfortunate consequence of this situation is that the intervening linkages, which are a prerequisite for any good theory, tend to slip from our grasp when we talk about social and economic differentials in mortality rates. Why do children born to poorly educated mothers have such poor prospects of surviving to adulthood? We think it is because they are more frequently exposed to disease-carrying pathogens. We believe as well that, should they chance to fall ill, such children will be less likely than others to benefit from some form of speedy and effective health care. These, of course, are plausible hyphotheses. But the actual evidence upholding them is generally weak or missing altogether.
The present study represents a response to this situation. Using national-level data from the 1993 National Demographic Syrvey (NDS), which canvassed approximately 13,700 Filipino mothers, we consider two of the most important and potentially lethal childhood ailments in the Philippines: diarrheal disease and acute respiratory infection (ARI). For each of these conditions, two separate questions may be asked: first, what factors influence the probability that a child will contract either diarrhea or ARI, and second, what factors determine the types of response (i.e., medical treatments) observed when a child does fall ill?
In delineating the major variables to be analyzed, we can perhaps begin with the second of the two study questions listed above. The theoretical underpinnings of such an analysis are simple and may be drawn from conventional theories of the diffusion of innovations (Rogers 1983). In general, we would expect mothers with greater access to information about modern health-care technologies (media users, the better educated, members of certain cultural or religious groups) to be more likely than others to follow a medically recommended treatment when their children fall ill. We can likewise hypothesize that parents with superior economic resources will be overrepresented among those choosing to use modern treatment options, particularly when those options are more expensive than traditional responses.
Another type of resource consists of parental availability to supervise children's treatment for illnesses. The children of working mothers and single parents, for example, may well fare poorly in this regard, although in the case of working mothers we can also hypothesize a counterbalancing effect due to increased household earnings. Geographic factors may also be important, as shown by comparisons among different regions of the country or among households located near or far from a health-care facility. Finally, it is possible that certain characteristics of the child (age, sex, birth order, whether the child's birth was desired or not) and of the morbid condition (severity of symptoms) will also play a role.
Many of the characteristics may be linked to the two serious childhood ailments of diarrheal disease and ARI. In general, we would expect that the social and economic variables most commonly associated with low rates of infant mortality--such as maternal education and urban residence--will also be correlated with low rates of these diseases (see Costello 1988 for a review of Philippine studies of infant mortality and its determinants). The argument here is that knowledge, economic resources, accessibility and the like should lead to an improved standing on the various "proximate" determinants (Mosley and Chen, 1984) of mortality, such as inadequate nutrition, poor housing housing, and lack of access to preventive health-care technologies, thereby reducing the chances of contracting a life-threatening disease.
Although the NDS data set on infant and child morbidity is extensive, previous analyses have been cursory at best. The final NDS report, for example, covers the issue of morbid conditions among infants and children with a seven-page segment from a single chapter (ROP, National Statistics Office, and Macro International 1994). We have therefore conceptualized the present research project as largely exploratory and cast a wide net in searching for potential determinants of morbidity and treatment.
But this, we hasten to add, will be a special type of exploration, one in which the early stages of the expedition have been purged from the final account of the journey, despite their obviously crucial character. For we have constraints of space and time to deal with which will not allow us to describe in loving detail the analyses of frequency distributions and bivariate cross-tabulations which comprised the bulk of our initial project reports. Instead we confine our attention largely to the multivariate results, for these are not only more definitive but also more succinct. In any event, the reader can be assured that the final model specifications did draw heavily from these earlier forays into the NDS landscape.
A major objective of the analysis is to provide insights for policymakers. Geographic comparisons should give us some idea of differential program impact, as separate from any underlying differences in developmental status that may already exist among regions or between rural and urban communities. In like fashion, the study should also serve to pinpoint those social and economic groups which are least able to protect the health of their children or to offer them an appropriate treatment once they fall ill.
Of particular interest, for both theoretical and policy-related purposes, is the set of child-specific factors associated with parental underinvestment theory (Scrimshaw, 1978). As originally formulated, this theory emphasizes the manner in which high rates of fertility in a society are linked to similarly excessive levels of infant mortality. The main idea is that frequent childbearing, as combined with the economic limitations experienced by most Third World families, militates against the full expenditure of family resources to protect the health of each newborn child. Resignation and apathy in the face of repeated morbidity experiences become, as it were, rational strategies for "investing" familial resources in a situation where there is an "oversupply" of young children.
An observer working wita a household-level data set might well hypothesize that unwanted or higher-parity children would be more likely to fall ill and less likely to receive adequate health carethan children who were fully desired by their parents and born into a small family. Scrimshaw notes that "in general, the distribution of food in a household may favor some individuals over others; when scarce resources must be distributed among several children, the more wanted children may receive more and better food than the others...[whereas] the less valued child is more likely to be taken to a health practitionerlater in the course of the illness, if at all" (1978, 394-95). She also makes specific mention of the potential link between short birth intervals and parental underinvestment.
A somewhat analogous issue has been raised by economists concerned with the allocation of resources within households. What we find again is an underlying assumption that parents will not always use all possible resources to protect the health of their children. On the one hand, some children (e.g., those less than a year old) may be perceived as weaker or more vulnerable and therefore worthy of additional care. On the other, discriminatory attitudes may lead parents to give some children less in the form of nutritious foods or good quality health care. A classic example of the latter pattern is found in the favoritism said to be exhibited toward sons in rural Bangladesh (e.g. Chen, Huq and D'Souza 1981). While there appears to be less evidence, on the face of it, that parental discrimination against girls exists in the Philippine, it is of interest interest such a possibility. Parents may also make health-care decisions on the basis of quite a rational criteria--for example, severity of symptoms exhibited by an ill child. Birth spacing and parenting skills also have clear relevance to the underinvestment thesis.
If parental underinvestment in children exists in the Philippines, it clearly has policy implications--especially now, given the recent shift in rationale for government-sponsored family planning programs from the promise of macro-level economic gains to the beneficial effects of birth spacing and family-size limitations upon maternal and child health. Pachauri (1995, 12), discussing the effects of a child's wantedness status (i.e., whether the mother had wanted to become pregnant at the time of the child's conception), favors a set of "reproductive health programs (that) would become responsible for reducing the burden of unplanned...child bearing and (its) related morbidity and mortality" (emphasis ours).
In this report we summarize what has become a detailed analysis of a rich and varied data set. One consequence of striving for brevity has been to forgo a conventional literature review, other than the few comments above. The balance of the report may therefore be divided into three roughly equal sections: (1) a discussion of data and methods, (2) the statistical presentation proper, and (3) a recapitulation of the study's major findings and policy implications. Readers interested in more details are invited to peruse our earlier reports (Costello and Lleno 1995a, 1995b, 1995c, 1995d, 1995e), which are available from the first author upon request.
The NDS survey module on infant and child morbidity comprises more than 70 items on such topics as disease prevalence, treatment patterns, preventive health practices (e.g., immunizations) and nutritional practices, particularly those pertaining to breastfeeding. In addition to questions about ARI and diarrhea, a question was asked about the occurrence of measles. The survey included a full gamut of potential morbidity and treatment determinants and may be said to represent four distinct levels of analysis: characteristics of the child (e.g., age, sex), of the mother (e.g., highest level of education completed), of the household (e.g., income, residential crowding), and of the local community (e.g., availability of health services).
Sample sizes for the present analysis will differ according to the particular dependent variable under consideration. In general, the dependent variables fall into three categories: (1) morbidity status, measured separately for ARI and diarrheal disease, (2) knowledge and use of oral rehydration therapy (ORT) as a curative resort for diarrhea, and (3) the type of health care worker, if any, resorted to during incidents of ARI and diarrheal disease.
The questions on infant and child morbidity referred to all children of survey respondents who had not yet reached their fifth birthday as of the survey date. From that group we eliminated all newborn infants (less than one month old) and all children living apart from their mothers, leaving a sample of 8,351 infants and children. Sample sizes for them and all other variables underwent an addtional (minor) reduction during the multivariate analysis: all cases taht had been scored with a missing value (no response, not applicable) on even a single predicator variable were automatically eliminated.
The ORT variables deal with the mother's knowledge and use of this therapeutic method and were therefore phrased with reference to the NDS respondents rather than their children. Sample sizes were 7,889 women for the questions on ORT knowledge and 6,837 dor the questions on ORT use. The second sample was smaller because women who knew nothing about ORT were not asked aboutn its use.
The data on consultations with health-care worker again referred to the infant and child data set. The sample size was greatly reduced because those questions were asked only with reference to children who had fallen ill with ARI or diarrhea during the two-week preceding the survey. The relevant sample sizes for this variable therefore stand at 1,721 for ARI and 838 for diarrhea cases. Because these samples were so small, we used the .10 level of probability (that is, a 10 percent margin of error) for statistical tests involving this particular dependent variable. All other analyses set alpha at .05 (5 percent).
The NDS questions on respiratory illnesses were composed of three items that inquired into the incidence of fever, cough and short or rapid breathing during the two-week period before the survey. Using a symptomatic definition of ARI provided by the World Health Organization (Cabaraban 1993, 9) we have determined the child's ARI status by combining all three indicators. All cases in which at least two such symptoms were present we designated as having contracted an acute respiratory infection. They accounted for nearly 21 percent of infants and children in the sample. This percentage probably represents a slight understatementof ARI prevalence because the question on short or rapid breathing was addressed only to mothers of children who had experienced a cough. It was therefore not possible to identify any case with the two-symptom combination of feevr and short or rapid breathing. (Marginal results for the three symptoms taken separately were 26 percent for fever, 33 percent for cough, and 9 percent for short or rapid breathing).
The presence of diarrheal disease was determined by a single question on whether the children in question "had diarrhea in the last two weeks." In all, 10 percent of the children under age 5 experienced a bout of diarrhea during the reference period. A question was also asked about the incidence of both cough and dairrheaduring the 24-hour period preceding the survey. In both cases the resulting levels were high enough to suggest that some cases of ARI and diarrhea that had occured toward the beginning of he two-week period were not recalled by respondents (Costello and Lleno, 1995d).
For the question on of knowledge of ORT, we coded as knowledgeable all women who affirmed either that they had heard about oral rehydration solution known as "ORESOL" or that they had ever seen ORESOL packet when shown one of these by the interviewer. Overall, 86 percent of the respondents said they knew about this type of therapy. The question on ORT use asked whether a respondents had ever "prepared" an ORS solution either for herself or for someone else "to treat diarrhea." Only 56 percent of all women had ever done so. As already mentioned, the population of women included in the statistical analysis of ORT use consisted solely of that subgroup of respondents who had ever heard of this remedy. In other words, the 44 percent who were not users of ORT all knew about the method but nevertheless had not used it. Those nonusers represented 52 percent of the total population of survey respondents. Some respondents reported using expensive and inappropriate medicines to treat diarrhea. For example, 17 percent of all children who were ill with dairrhea during the two-week period prior to the survey had been given antibiotics.
The third dependent variable concerned health specialists. We combined responses to an initial question about whether the respondent had obtained any "advice or treatment" with responses to a follow-up question on the particular palce or practitioner consulted. Responses to this second question were assigned to nine major categories. These were treatment at or by (1) a government hospital or clinic, (2) a rural health unit, (3) a barangay (community) health station, (4) a private hospital or clinic, (5) a private physician, (6) a community health worker, (7) a pharmacy, (8) a traditional health-care worker (e.g. hilot), and (9) all other responses. While the nine options could be grouped in several ways, we chose the following three-category typology:
1. cases referred directly to a physician or to a health center with an affiliated physician (subcategories 1, 2, 4 and 5 above);
2. Cases referred to all other health-care workers (subcategories 3,6,7,8 and 9);
3. Cases not referred to any health-care worker.
Among all cases of children suffering from ARI, a total of 32 percent had visited a physician or a health center with an affiliated physician (henceforth referred simply as a physician), 28 percent seen some other practitioner, and 40 percent had not seen any health-care worker. Corresponding proportions for diarrheal diseases were 23 percent, 22 percent and 55 percent, respectively.
Exhibit 1 presents the cursory operational definitions of the predictor variables. Readers interested in knowing more about the distributional aspects of these factors are referred to our initial project report (Costello and Lleno 1995d).
Despite the wide range of variables included in the multivariate analysis, some observers may as yet note the absence of still other factors relevant for the study of morbidity conditions and their treatment. Major possibilities include ethnicity, maternal literacy status, the use of breastfeeding and supplementary foods, and various community-level variables (e.g., type of water system, presence of a sewer system, type of access road to the barangay). These variables were eliminated from the final models because they would have led to an unacceptable reduction in sample size (the nutritional indicators and the community-level variables, for example, were elicited only within rural barangays), because their essential qualities were captured by some other highly correlated indicator (e.g., maternal literacy and highest level of educationcompleted by the mother), or because the initial analysis indicated the possibility of data quality problems. The last outcome was true for a number of the community-level factors.
Of particular interest here in the child's breatfeeding status. Because the question on this topic was asked only about the youngest child, to include responses would have resulted to an initial reduction in sample size. The relationship obtained at the bivariate level indicated that the effect of this factor depended heavily upon the use or nonuse of supplementary foods. Breastfeeding in itself was not associated with a reduction in the incidence of either ARI or diarrheal disease; if anything, the opposite tendency held. However, the minority of babies who weer not given any supplementary food did show a lower morbidity level for both conditions. adding the question on supplementary feeding to the final model would have, again, greatly reduced the sample size because it was asked only with reference to breastfed children.
Given the categorical nature of each of the study's dependent variables, logistic regression ("logit") was an appropriate statistical tool for a multivariate analysis. In the case of morbidity status and ORT acceptance this involved a binary choice model, whereas the study of health-care treatment will require the use of multinomial logit regression.
Because the logistic function deals with estimated probabilities, our multivariate (binary) models can be written as:
logit P = log P/1-P = b0 + b1 X1 + b2 X2 + ... + bk Xk + e,
where P is the probability, say, of falling ill during the reference period and X1, X2, etc., represent the various predictor variables included in the model. The resulting logit coefficients measure the effects of these factors and may be interpreted initially in terms of their sign (whether positive or negative) and statistical significance. Unfortunately, the absolute values of the logit coefficients are not easily interpreted. For this reason, we carried the analysis one step further by computing the adjusted probabilities of becoming ill (Retherford and Choe 1993, chapter 5). The procedure is based on a multiple classification analysis (MCA) of the dependent variable that yields the probability that an infant or child belonging to a certain social category (e.g., residents of an urban barangay or those aged exactly 1 year old) will fall ill once all other factors in the model have been held constant. We therefore computed three sets of statistics for each model--its associated logit coefficients, their significance levels, and the adjusted probabilities of falling ill (or of a mother taking an ill child for treatment or knowing about or using ORT)--for all variables found to be statistically significant. We will not report the first of these three parameters here; more detailed information on statistical regressions is available in earlier publications (Costello and Lleno 1995a, 1995c).
In Table 1 the factors (predictor variables) significantly correlated witha health outcome are indicated with their corresponding levels of statistical significance; those not significantly correlated are designated by "ns" (not significant). Dashes are used to show all cases in all cases in which a variable was excluded from the model. The table may therefore be read horizontally to see the overall effect of each variable on the morbidity and treatment factors considered as a whole. Conversely, the table can be read vertically to indicate the outcomes produced by each individual model. All the models were significant as a whole at well below the .001 level.
The nature of each significant association may be understood through two devices. First, in Table 1, an asterisk designates each significant relationship that operates in a manner contrary to what might conventionally expected. as an illustration, adjusted morbidity levels for ARI and diarrhea (both marked with an asterisk) tend to be higher in urban than in rural barangays. By contrast, the significant association between rural/urban residence and diarrheal treatment indicates that urbanites are more likely to treat infant or child diarrhea with a visit to a physician. Because this latter finding accords with our expectation there is no asterisk. Second, by showing the adjusted percentages for all significantly correlated variables and thereby showing clearly the magnitude of all intracategory differences, Table 2, 3 and 4 provide a more precise description of these relationships. For the rural-urban comparison, for example, we see that rural-based households have a slightly lower incidence of diarrheal disease (8 percent compared with 10 percent) once all other factors are held constant.
The findings are arranged according to the role played by each predictor variable in affecting the morbidity and treatment factors. Our discussion follows the sequence of variables in the list of definitions (Exhibit 1). In general, we begin with community-level factors, then move on to maternal and household variables, and end with characteristics associated with children.
Rural-urban residence. Residence in a rural or urban area was not significantly related to either of the ORT variables of to the treatment of children with ARI. As mentioned earlier, rates of both ARI and diarrheal disease were moderately higher in urban barangays. This, of course, was for the adjusted figures, the relationship observed when all other factors were held constant. In comparison, the bivariate (unadjusted) statistics showed an opposite tendency: higher illness rates in the rural barrios. Apparently rural households experienced greater overall morbidity, but this could be explained entirely by their lower levels of such factors as education and income. In comparison, the "pure" effect of rural living in a rural area was basically positive.
Urban children with diarrhea were more likely than rural youngsters to be brought to a doctor but less likely to go to any other health-care practitioner. Rural-urban differences in cases not brought to any practitioner were slight.
Region. Region of residence is associated with each of the six dependent variables at the .001 level of significance (1 percent margin of error). Its overall importance as a predictor variable is therefore clear. Unfortunately, the precise manner in which this factor affected both morbidity and treatment is less easily summarized.
Let us begin by terming Metro Manila, Central Luzon and Southern Tagalog (possibly including the Central Visayas) as the "core" regions of the country. Adjusted morbidity rates for both diseases were generally low in that area than elsewhere (Table 2), as were the knowledge and use of ORT (Table 3). (Note, however, that the latter finding did not in Metro Manila but hold for Central Visayas). Physician use was consistently high somewhat below average in the other three regions, especially for diarrhea (Table 4).
Cagayan Valley, Bicol, Western Visayas and Eastern Visayas are generally considered to rank among the poorest regions in the country. As a whole these locales had above-average morbidity levels along with somewhat greater knowledge about ORT. The other three analyses failed to reveal ant clear pattern for this group. The surprisingly high levels of physician use in the Western and Eastern Visayas are of some interest. While they may bode well for the health of children in those regions, the very low use of other health-care workers in Western Visayans made the region's ranking in the use of health-care services among the lowest of all regions.
For some reason, the Cordillera Autonomous Region rated particularly high on knowledge and use of ORT. Northern Mindanao exhibited a very high use of nonphysician health services (probably including the barangay health station), and as a result it ranked among the highest in the use of health-care services.
Community electrification. Electrification was included in only the two models of health-care services utilization. For diarrhea the relationship was nonsignificant, but for ARI the physician use rates were higher (32 percent) in communities with electricity (32 percent) than those without (22 percent).
Community toilet facilities. It might be expected that poor sanitation--as indexed by a high proportion of households with inadequate toilet facilities--would increase the incidence of diarrheal disease. However, this prediction was not borne out by the data.
Distance to a health facility. Are parents who live far from a clinic less likely to seek treatment for their sick children? We found weak (p<0.10) evidence to this effect for ARI but none at all for diarrhea.
Religion. Comparisons between Roman Catholics, Protestants, and the residual "other" category failed to show any striking differences. Members of the local Christian sects Iglesia ni Kristo and the Aglipayanism seemed somewhat more favorable to ORT and the use of "other" health-care providers. They also showed evidence of higher levels of infant and child diarrhea.
Muslim respondents scored lower than other religious groups on both the knowledge and use of ORT. It is evident that the message to use ORESOL is not getting out to members of this group. Muslim parents were less likely to bring a child with ARI to any health-care worker.
Marital status. Single parent families (in which the spouse was absent because of death, legal separation or employment outside the community) did not not show evidence of higher levels of infant or child morbidity than other families. Nor did this factor have any discernable effect on the knowledge or us of ORT. Marital status had to dropped from the model of health-care service utilization because the number of single-parent households was too small for the results to be statistically reliable.
Mother's age. Maternal age was included in all six models. In no case was a statistically significant relationship obtained.
Children ever born. We had expected that high-parity children--those born to mothers who had already borne a large number of children--might be more prone to illness than other children, but the data showed no evidence of such effect. Nor were high-parity children any less likely than other children receive medical care once they fell ill.
Both knowledge and use of ORT were positively associated with the number of children ever born. The reason for this is finding is self-evident: mothers with more children tend to have more exposure to childhood diarrhea and thus more opportunities to learn about and use ORESOL.
Mother's education. The higher the educational attainment of a child's mother, the greater was the probability that the child had not had ARI during the two-week period before the NDS. In the case of diarrhea, however, no significant relationship emerged.
Maternal educational was associated with knowledge, but not use, of ORT. For children with ARI, maternal education was associated positively with the use of another type of health-care practitioner. for children with diarrhea, maternal education had no significant bearing on the use of health services.
Mother's work status. Mother's work status was not be significantly related to the incidence of childhood diarrhea. The lowest level of ARI were found among the children of white-collar workers and women not in the labor force (housemakers). Mothers engaged in farming or fishing were somewhat less likely than others to have used ORT. As might be expected, this group was also less likely tha others to bring a child with ARI to a physician.
What is perhaps most interesting for the health-care treatment issue, though, is the pattern orserved for mothers in the white-collar occupations. For ARI, we found moderately high level of physician use coupled with an extremely infrequent use of all other health-care workers. As a result, this group ranked lowest in the provision of professional health treatment to children with ARI. To illustrate, 55 percent of the children with ARI whose mothers were white-collar workers were not brought to any health-care worker, compared with 39 percent of of children whose mothers were service or blue-collars. A tentative explanation is that women working in formal-sector jobs do not have the flexibility to take care of an "unscheduled" problem such as a sick child. No significant relationship was obtained for diarrheal treatment, however.
Father's education. Paternal education was not significantly related to either of the morbidity-status measures or to knowledge and use of ORT. In the case of their children with ARI whose fathers were poorly educated, though, the expected pattern prevailed; they were less likely to be brought to any health-care worker, particularly a physician. For children with diarrhea, a somewhat different pattern emerged. Ill children of less-educated fathers were less likely than those of more educated fathers to be taken to a physician but more likely to be taken to another health-care worker. Indeed, children of fathers who never went to school are four times more likely to be brought to a health-care worker other than a doctor than were children whose fathers had a college degree. As a result, it is the better-educated fathers were the least likely group to be brought to any health-care if they contracted diarrhea.
Father's occupational status. Disease rates tend to be highest among children with fathers employed in agriculture or fishing. This pattern holds both for ARI and diarrheal disease. Women married to white collar workers (here defined as men holding professional, clerical, or sales positions) were generally less knowledgeable about ORT than women whose husbands were in other occupations; they were also less likely willing to use ORT than women whose husbands were in other occupations. A somewhat similar pattern prevailed for health-care treatment: for children with ARI, the wives of white-collar workers were less likely than other women to consult a health-care worker apart from a physician. As a result, the wives of blue-collar workers are more likely than the wives of white-collar workers to bring a child to any health-care worker.
Ownership of consumer items. The ownership scale was not significantly correlated with either of the two measures of morbidity status. As would be expected, wealthy families tended to take their children to a physician, at least for cases of diarrhea. Wealthy mothers were somewhat less likely to know about ORT than poor mothers and were much less likely to have used ORESOL.
Housing quality and housing density. Housing quality was originally seen as a Mosley-and Chen-type proximate factor that ought to mediate the inverse relationship between social class and morbidity levels. After controlling for all other factors in the model, however, it was not significantly associated with the occurrence of either ARI or diarrheal disease. The data also failed to substantiate an essentially similar hypothesis related to household density.
Media use. The assumptionto that media exposure would increase maternal health knowledge and, through this, reduce infant and child morbidity levels was not borne out by the data. The relatioship for ARI was not significant, whereas the incidence of childhood diarrhea tended to be higher, not lower, as media use went up results for the treatment of ARI were roughly analogous: women who scored lowest on the media index were mostly likely to bring a child to either a doctor or another type of health-care worker. By contrast, we found a positive association between media exposure and knowledge and use of ORT.
Child's age. The relationship between child's age and the incidence of illness was curvilinear. Illness levels were moderately high and rising during the first year of life. They reach a peak at about 18 months thereafter declined. The younger an ill child was, the greater was the likelihood that he or she would be taken to a physician. In the case of ARI, older children were less than younger children to visit any type of health-care worker.
Child's size at birth. Birth size was considered for only the two morbidity models. As expected, children with low birth weights were more likely than those with large or average birth weights to fall ill with ARI. For diarrheal disease, the relationship was significant.
Child's sex. All comparisons for this factor were statistically nonsignificant. We could discover no evidence of discrimination against daughters. In fact, the logit coefficient for consultation with a physician in cases of diarrhea showed a somewhat lower utilization level for sons: the relationship barely missed significance at the .10 level of probability.
Length of preceding birth interval. Longer birth intervals did not found have any significant impact upon morbidity levels. They were, however, associated with better health care treatment, at least in the case of ARI. For example, 61 percent of children with ARI who were born after an interval of four years were brought to a physician or other health-care worker, as compared with 55 percent of children born after a one year interval.
Wanted status of birth. Three categories of wantedness were compared: children who had been a desired from virtually the time of their conception, children who were accepted by their mothera as "a desired birth" at some later date, and children who were still considered unwanted at the time of the survey. The last group is, of course, the key one. It included 16 percent of all children.
The four comparisons for this variable approximated the predicted direction, and the incidence of and treatment for ARI statistical significancant. Thus, 24 percent of not-wanted children had fallen ill with ARI during the two-weeks before the survey compared to 18 percent of the "always wanted" group. Similarly, 49 percent of not-wanted children with ARI were not treated by any health-care worker, compared with 39 percent of those who had always been wanted.
Severity of Symptoms. The greater the severity of the child's symptoms, the more likely was he or she to have been brought to a doctor. This pattern holds true for both diseases at the .01 level or lower. The severity of symptoms did not affect the likelihood of a child seeing another type of health-care practitioner, however.
We began this inquiry in the hope that the study of morbidity and treatment differentials might lead us to a better understanding of why certain social and economic groups in the Philippines have such high rates of infant and child mortality. Have these expectations been justified?
The study also estimated levels of health-care utilization. Overall, 52 percent of children who had suffered from acute respiratory infection during the two-weeks before the survey had seen a doctor or other health-care professional. Of the children who had suffered diarrhea, 45 percent had seen a health-care professional, and 18 percent had received oral rehydration therapy. Clearly, the Philippine Department of Health needs to continue to extend its services, despite laudable gains already made.
Some hypothetical linkages between childhood illness and general levels of poverty were not borne out by the NDS data. This was true, for example, of the indicators of housing quality, household density, community toilet facilities. Although the are undoubtedly many reasons why the country should continue to press for improved housing and sanitation, further improvements in those areas may not have an immediate effect on the health status of Filipino children. The provision of additional health clinics must also be rated somewhat minor polci recommendation at this point, given the weak correlation found between the treatment of children and the household's distance from the nearest clinic.
The study did not establish any convincing linkage between either morbidity status or health-care provision and the sex of the index child. Earlier analyses also failed to show much, if any, evidence that Filipino girls were more likely to die in infancy than boys. Without discounting the importance of gender inequalities in Philippine society, we therefore have to agree with Herrin (1994) that this issue does not appear to have serious consequences for the health-care of female infants and young girls.
We found some interesting morbidity differences between rural and urban barangays. For both ARI and diarrheal disease, the multivariate analysis and the bivariate analysis revealed opposite patterns. Initial tabulations showed fewer cases of ARI and diarrheal disease, on average, in the cities than in rural areas. After we controlled for such correlated factors as education, occupational status, however, these findings were reversed. Exactly parallel results were obtained by one of the best-known multivariate analyses of infant mortality ever conducted in the Philippines (Martin et al., 1983). These findings raise disturbing questions about living conditions Philippine cities (see also Cimatu 1995) and suggest that effects of "economic development" on the health of Philippine infants and children may not be entirely positive. Current rapid rates of urbanization lend special emphasis to these concerns; census data show that the proportion of Filipinos living in cities grew from 37 percent in 1980 to 49 percent in 1990 (ROP, National Statistics Office 1992, table A).
Another concern is the rapidly growing number of working mothers and single-parent families, due chiefly to structural and economic changes associated with the development process. Are such changes adversely affecting child welfare? Not according to our findings. Children living in one-parent households were more likely to fall ill than children living with both their parents. For diarrheal disease, maternal work status was not significantly associated with either the probability of a child's falling ill or the type of treatment provided. Results for ARI were mixed, although we did discover a surprising pattern: mothers in white-collar and professional occupations were less likely than other mothers to consult a health-care worker when their children suffered from respiratory infection. We speculate that women women in those occupations may experience particularly severe time constraints.
Increased use of the mass media represents a second institutional change associated with economic development. Our findings here were unexpected. Exposure to media was associated with better knowledge and use of oral rehydration therapy but was not associated with lower child morbidity or increased use of health-care practitioners. In fact, the data revealed the opposite pattern for diarrheal morbidity and treatment of acute respiratory infections.
Both the Philippine Department of Health and the World Health Organization recommended the use of oral rehydration therapy to treat diarrheal disease. In the Philippines, ORESOL tablets are available without a prescription and have been given fairly wide publicity by the Philippine Department of Health under the Primary Health Care Program. Nevertheless, our study revealed that a number of Filipino families were treating childhood diarrhea with expensive, and someyimes inappropriate, medicines such as antibiotics.
It is a sociological truism that culturally innovative behaviors usually begin in the higher, more urban social classes and diffuse downward to lower-class and rural households (Sorokin 1959). This does not appear to be the pattern, however, for oral rehydration therapy in the Philippines. What we found instead were several instances of an inverse relationship between the respondent's knowledge and use of ORT and her social or economic status. This was true, for example, for the scale of consumer goods owned and several of the regional comparisons. Results were similar for treatment by nonphysicians, a variable that provides an approximate measure of the public's willingness to use barangay health stations.
To some extent these pattern are to be expected. After all, if parents can afford to bring a child to a private-sector physician, why should they bother to go to the local health station? Nonetheless, these findings do raise several concerns. Some of our multivariate analyses, for example, showed that high-status parents were less likely to take their ill children to any health-care practitioner, presumably because of their disdain for treatment by nonphysician. Yet in some cases greater use of government services might improve the health of children from wealthy families.
We also know from survey evidence that many lower-class mothers do not place much faith in the quality of services offered by the Department of Health in general and by the barangay health stations in particular (Costello and Palabrica-Costello 1994; Palma-Sealza 1993). Perhaps they have observed that better-educated and higher-status parents are not using the government facilities and therefore have deduced that the services is of poor quality. To view the problem from a slightly different perspective, might not the quality of care improve if the Department of Health clinics could attract at least some higer-status clientele? The logic here is that such persons might be able to insist on better service, improved technologies, and more courteous attitude from the staff.
We found substantial gaps between the proportion of respondents who had ever heard of oral rehydration therapy (48 percent), who had ever used ORT (48 percent), and who hade used it during the most recent diarrheal episode (18 percent) (Costello and Lleno 1995a, table 8). A major challenge for the Department of Health is to convince people taht ORESOL is effective and to persuade them, when a child develops diarrhea, to use the tablets quickly and consistently.
The strong--perhaps too strong--association between ORT and the government health service may be backfiring. When NDS respondents were asked if they knew of a person or a place where ORESOL could be obtained, the three most frequently mentioned locales were all connected with the department of Health: barangay health stations (46 percent), rural health units (26 percent), and government-run hospitals (14 percent). In contrast, only 5 percent of respondents mentioned either a private physician or a private hospital. Members of the public may be making the unwarranted assumption that ORT is second-rate therapy suitable only for the poorest families. Additional efforts are needed to move ORT into consumer-marketing mainstream, for instance through the production and adverstising of privately manufactured brands (possibly under a social-marketing program) and through a campaign to encourage private-sector physician to describe ORESOL tablets more often.
Suprisingly, Metro Manila ranked lowest ranked lowest of all 14 rehions on the level of ORT knowledge. Data from the 1987 National Health Survey, as reported by Herrin et al. (1993, figures 2.10a and 2.10b), show essentially the same pattern. Until now, the Department of Health appears to have concentrated its ORT dissemination efforts in the more peripheral, rural areas of the country. If that is the case, we suggest that the time has come to redress the balance, particularly given the continued movement of Filipinos to the National Capital Region (Costello and Ferrer, 1993). Stronger efforts to encourage ORT use among Muslim Filipinos are also in order.
Our findings on the use of ORT and nonphysician health services aslo suggest that the conceptual dichotomy between "modern" and "traditional" health-care services no longer suffices for the Philippines. In its place we offer a threefold typology that distinguishes traditional services from two types of modern services. The first involves private sector clinics and hospitals, physician-specialists and expensive medicines produced by major pharmaceutical companies. The second comprises all other types of modern treatment--those associated with the Department of Health, with paramedical health-care workers, and with low-cost type of modern treatment such as ORESOL. More affluent Filipinos tend to use the first type of modern service, even in atypical cases in which there are good scientific reasons for doubting its validity (such as use of antibiotics for diarrheal therapy). For their part, poorer Filipinos must content themselves with the second category of modern treatment or with traditional treatment, although it appears that the purely traditional techniques are now being used less and less frequently: fewer than 4 percent of the children who were ill with either ARI or diarrheal disease were brought to a traditional healer.
Our final comment brings us back to our earlier speculations on parental underinvestment theory. In general, we found support for this theory. Family size consistently failed to predict either a child's morbidity status or health-care treatment, but children born after longer birth intervals were being given better treatment, at least for ARI, than those who followed soon after the next-elder sibling. And the wantedness factor produced even stronger results: unwanted children were not only being more likely to fall ill with ARI but also less likely to receive adequate health than wanted children. The NDS findings thus tend to validate the slogan "Family planning saves lives" and they give pratical meaning to the statement "Every child [should be] a wanted child."
Cabaraban, Magdalena C. 1993. Rural Mother's Management of Acute Respiratory Infection (ARI) in Region 10. Project Report to the Health Research Network. Cagayan de Oro: Research Institute for Mindanao Culture, Xavier University.
Chen, Lincoln, E. Huq. and S. D'Souza. 1981. Sex Bias in the Family Allocation of Food and Health Care in Rural Bangladesh. Population and Development Review 7:55-70.
Cimatu, Frank. 1995. WHO Study: Manila Pollution among the Worst. Philippine Daily Inquirer 10:328 (Nov. 2), p. 17.
Costello, Michael A. 1988. Infant and Childhood Mortality Research in the Philippines: Review and Agenda. Journal of Philippine Development 15(2):257-274.
Costello, Michael and Pilar L. Ferrer. 1993. Recent Trends in Population Redistribution in the Philippines with Special Reference to Metropolitan Manila. Paper presented at the 22nd General Population Conference of the IUSSP, Montreal.
Costello, Michael and Lauro C. Lleno. 1995d. Infant and Child Morbidity in the Philippines: A Profile Analysis from the 1993 National Demographic Survey. Report No. 1 to the NDS-SMS Secondary Analysis Project. Cagayan de Oro: Research Institute for Mindanao Culture, Xavier University.
_______________ . 1995b. Differential Morbidity among Infants and Children in the Philippines: An Exploratory Analysis. Report No. 2 to the NDS-SMS Secondary Analysis Project. Cagayan de Oro: Research Institute for Mindanao Culture, Xavier University.
_______________ . 1995e. Social and Economic Differentials in Treatment Patterns for Infant and Childhood Disease: Philippines, 1993. Report No. 3 to the NDS-SMS Secondary Analysis Project. Cagayan de Oro: Research Institute for Mindanao Culture, Xavier University.
_______________ . 1995a. Differential Morbidity among Infants and Children in the Philippines: A Multivariate Analysis. Report No. 4 to the NDS-SMS Secondary Analysis Project. Cagayan de Oro: Research Institute for Mindanao Culture, Xavier University.
_______________ . 1995c. Differential Treatment Patterns for Acute Respiratory Infections and DiarrhealDisease: A Multivariate Analysis of Recent Philippine Data. Report No. 5 to the NDS-SMS Secondary Analysis Project. Cagayan de Oro: Research Institute for Mindanao Culture, Xavier University.
Costello, Michael A., and Marilou Palabrica-Costello. 1994. Health Care Financing for the Rural Poor: Findings from a Survey Study. Final project report to the DOH-PIDS Project on Health Care Financing Reform. Cagayan de Oro: Research Institute for Mindanao Culture, Xavier University.
El-Zanaty, Fatima, Hussein A Sayed, Hassan H. Zaky, and Ann A. Way. 1993. Egypt Demographic and Health Survey 1992. Calverton, Maryland: Macro International, Inc. Etemadi, Felisa U. 1995. Urban Poverty Alleviation in Cebu City. Paper presented at the ASEAN Inter-University Seminar on Social Development, Cebu City. 28-30 November.
Flieger, Wilhelm, Macrina K. Abenoja, and Alice C. Lim. 1981. On the Road to Longevity: 1970 National, Regional and Provincial Mortality Estimates for the Philippines. Cebu City: San Carlos Publications.
Herrin, Alejandro N. 1994. Population, Health and Women. Journal of Philippine Development 21 (1 and 2): 341-382.
Herrin, Alejandro N., Aleli de la Paz Kraft, Oscar F. Picazo, Orville C. Solon, Mario M. Taguiwalo and Ma. Socorro Zingapan. 1993. Health Sector Review: Philippines. HFDP Monograph No. 3. Manila: Health Finance Development Project.
Martin, L.G., J. Trussel. F. Reyes-Salvail and N.M. Shah. 1983. Covariates of Child Mortality in the Philippines, Indonesia and Pakistan: An Analysis Based on Hazard Models. Population Studies 37:417-432.
Mosley, W. Henry and Lincoln C. Chen. 1984. An Analytical Framework for the Study of Child Survival in Developing Countries. Pp. 25-48 in W. Henry Mosley and Lincoln C. Chen (eds.), Child Survival: Strategies for Research. Population and Development Review (Supplement to Vol. 10).
Pachauri, Saroj. 1995. Defining a reproductive health package for India: A proposed framework. South and East Asia Regional Working Paper 4. New Delhi: Population Council.
Palma-Sealza, Lita 1993. Factors Affecting Utilization of Barangay Health Centers: A Survey Study. Project report submitted tothe Health Research Network. Cagayan de Oro: Research Institute for Mindanao Culture, Xavier University.
Republic of the Philippines, National Statistics Office. 1992. Socio-economic and demographic characteristics, Philippines. 1990 Census of Population and Housing: Report 3. Manila: National Statistics Office.
Republic of the Philippines, National Statistics Office, and Macro International. 1994. Republic of the Philippines: National Demographic Survey, 1993. Calvetron, Maryland: National Statistics Office, and Macro International.
Retherford, Robert D. and Minja Kim Choe. 1993. Statistical Models for Causal Analysis. New York: John Wiley and Sons.
Rogers, Everett M. 1983. Diffusion of Innovations. Third Edition. New York: The Free Press.
Scrimshaw, Susan W.M. 1978. Infant Mortality and Behavior in the Regulation of Family Size. Population and Development Review 4:383-404.
Sorokin, Pitirin A. 1959. Social and Cultural Mobility. Glencoe, Ill.: The Fress Press.