The rising occurrence of cesarean section has attracted international attention.1–6 In Hong Kong, 27% of all births were surgical deliveries in 1999. The mean cesarean section incidence in public hospitals (with a more serious case-mix and 74% of deliveries) was 19% and the corresponding rate in the private sector was 50%. This public-private gap widened by 64% during the 1990s and has continued.7,8 Case-mix and clinical indications have had little direct influence on similarly rising cesarean trends in the United States.9
To date, no study has taken into consideration the full range of health outcomes after cesarean section in an unselected population. We therefore examined the patterns of doctor visits and hospitalizations, overall and categorized by diagnostic groups, in a population-based birth cohort study of Hong Kong Chinese infants during the first 18 months of life.
Sources of Data
Data were drawn from a prospective birth cohort study conducted at all 47 Maternal and Child Health Centers of the Hong Kong Department of Health in 1997. For the index year, 92% of all infants born in Hong Kong were seen at one of these health centers at least once. The sampling frame consisted of all infants born in April and May of 1997 and brought to the health center for their first postnatal visit within 2 weeks after birth. The response rate to our study was 95%, comprising 88% of all births in the period. Mothers provided information on method of delivery, infant feeding history, household smoking habits, and other demographic, obstetric, behavioral, and potential confounding variables via a standardized self-administered questionnaire at their baseline health center visit. In particular, method of delivery was categorized as normal vaginal birth without instrumentation, vaginal birth with suction or forceps, or cesarean section. For the analysis, we collapsed all vaginal births into a single reference category (unassisted and assisted). We did not collect information on the possible indications of surgical or instrumental delivery, or whether the cesareans were primary versus repeat procedures.
Follow-up questionnaires were distributed at each subsequent routine well-baby check visit (baseline, and 3, and 9 months) to record health services utilization and to update other information (eg, breast-feeding patterns) for each intervisit interval (baseline to 3 months, 3 to 9 months, and 9 to 18 months). Utilization of health services was measured by the number of physician consultations and episodes of hospitalizations since the last follow-up, excluding preventive visits such as vaccination and regular check-ups. Mothers (or fathers, depending on who was the main caregiver) were asked about the type of illness leading to the consultation or hospitalization (grouped as febrile illnesses, respiratory illnesses, gastrointestinal illnesses, skin conditions, and all other conditions; all illness episodes consisting of fever, regardless of underlying cause, were categorized as “febrile illnesses” and thus the rest of the categories excluded concurrent febrile illness). To minimize recall bias, parents were given the health services utilization questionnaires to take home and were asked to fill in the form whenever the infant attended outpatient consultations or was hospitalized. In this way, the outcomes of interest were documented prospectively and recorded as they occurred. Using telephone interviews, trained research assistants contacted those who could not complete the questionnaire and those who did not return the questionnaire on schedule.
Informed consent was sought from all adult participants (on behalf of the infants) before enrollment. The project received ethics approval from the Research Ethics Committee, Faculty of Medicine, The University of Hong Kong.
We restricted the analysis to singletons born at term (ie, gestational age between 37 and 41 weeks comprising 84% of the sample) and whose mothers had never smoked (95%). This was done because multiple births, preterm (or post dates) births, and smoking were rare events, and thus exclusion was an efficient way to control for these conditions. Further, we included in the final analysis only parent–infant pairs with at least one follow-up visit and with no missing data in any of the data fields used in the multivariate modeling. To ensure representativeness of this final sample of study subjects, we compared those included with those excluded using Pearson's χ2 tests.
According to routine aggregate statistics in 2004, average hospital stay after cesarean section was 7.9 days, compared with 3.6 days after vaginal delivery. This difference creates a potential bias in that health care episodes in the immediate neonatal period for the vaginally born would have presented as “extra health care use,” while the surgically born who suffered from the same conditions would have been cared for as part of the initial birth hospitalization. We therefore discounted all health care episodes reported on or before 8 days after birth.
We employed bivariable and multivariable logistic regression analyses to study the association of method of delivery with doctor consultations and with hospitalizations of infants, as cumulatively recorded during follow-up visits at around 3, 9, and 18 months of age. Doctor consultations were dichotomized at the median number of visits, adjusted pro rata for length of follow-up where return visits differed slightly among individuals. This was done for all illnesses, as well as for each type of diagnosis. Hospitalizations were similarly coded; the median number of episodes was zero and therefore the outcome measure was dichotomized as ever versus never having been admitted to hospital. This dichotomy assumed that all infants were expected to incur some health services use; our goal was to capture the differential risk associated with method of delivery.
For the multivariable analyses, potential predictors were identified based on known confounders documented previously.10–12 Models were built hierarchically: Model 1 was unadjusted for any covariables; Model 2 was adjusted for the 2 covariables representing socioeconomic position, namely type (private vs. public) of hospital during delivery and highest parental education; Model 3 was additionally adjusted for maternal and infant characteristics (maternal age, gestational age, birth order, sex of infant, birth weight, and utilization of outpatient visits or hospital admissions where appropriate) plus the covariable indicating substitution of different forms of care; and finally Model 4 was additionally adjusted for mutable covariables, including exposure to environmental tobacco smoke after pregnancy, environmental tobacco smoke during pregnancy and breast-feeding history. To separate the effects of method of delivery on hospitalization from effects on outpatient doctor consultation, models 3 and 4 for utilization of inpatient hospital services controlled for the utilization status of outpatient visits (higher or lower) and vice versa. A priori, one might expect that there was a certain substitution effect between one form of care and another, especially in the local setting where public inpatient care is virtually free to all residents while private outpatient care is mostly fee-for-service.
We computed adjusted odds ratios (ORs) and 95% confidence intervals (CIs), using Stata version 9.2 (StataCorp, College Station, TX).
There were 8327 mother–infant pairs in the baseline sample. Of these, 4948 (59%) attended all 3 scheduled follow-up visits, 2114 (25%) attended 2 of the 3 follow-ups, and 828 (10%) attended only 1 follow-up. There were 437 (5%) infants who never returned after the baseline visit and thus provided no data for analysis. Restricting the analysis to term (between 37 and 41 weeks gestation) singletons and those whose mothers never smoked, and further excluding cases with missing values for any of the model parameters (n = 490), the final sample size for multivariable analysis was 5449 subjects. Comparing subjects who were excluded or missing with those included showed no substantial differences in method of delivery, use of outpatient or hospital services, or other measured variables.
Table 1 shows the distribution of variables according to baseline parent–infant characteristics and the associated crude odds ratios for doctor consultations and hospital admissions. More than one-quarter of all births were delivered by cesarean section and a further 17% were delivered vaginally with instrumentation; the remaining 56% infants had a normal, unassisted vaginal birth. Infants delivered by cesarean had a higher unadjusted risk of incurring more outpatient visits than average but were less likely to have been ever admitted into hospital from birth to 18 months overall (Table 1). The same pattern of association was observed for type of hospital for the delivery episode and highest parental education attained. Those who had a surgical delivery were much more likely to have been born in a private hospital (cesarean rate of 47% in private vs. 17% in public hospitals) and their parents were better educated (37% of parents with a cesarean baby vs. 26% of those with babies born vaginally had attained at least grade 12 education), suggesting confounding of the relation between birth method and health care use by socioeconomic position. Specifically, better educated parents were either more inclined to seek outpatient care, or had better access to such, or both. On the other hand, babies born to more highly educated parents may be healthier and therefore less likely to require inpatient care, which is less discretionary compared with ambulatory episodes and thus more reflective of actual health care need.
By controlling for the 2 indicators of socioeconomic position (type of hospital at birth and highest parental education attainment) in Model 2, the adjusted ORs regressed toward unity for both inpatient and outpatient use (Table 2). Further statistical adjustment in Models 3 and 4 did not change the point estimates or confidence intervals substantially. Adjustment of inpatient visits by the number of outpatient visits and vice versa did not change any of the ORs appreciably (data not shown). The results were also robust to separately considering unassisted (reference category) and instrument-assisted vaginal births.
After further disaggregating the fully adjusted results (Model 4) by parent–reported diagnostic groups (Table 3), surgical delivery was associated with more outpatient care use for afebrile gastrointestinal illness and skin conditions in the first 3 months. Higher outpatient utilization was also observed for afebrile respiratory and other conditions, as was an increased risk of inpatient care for afebrile gastrointestinal illness from 3 to 9 months. However, none of the time periods (ie, 0–3, 3–9, 9–18 months) appeared to be any more susceptible in general.
Furthermore, we did not see any consistent association between the use of inpatient or outpatient health services and the type of vaginal delivery (assisted with forceps/suction or no instrumentation).
Cesarean section is indicated for difficult operative vaginal and breech deliveries1,13,14 as well as prior cesarean births.15 However, changes in these indications are not enough to account for most of the rise in surgical birth rates in recent years.9 There have been sporadic reports that link cesarean delivery to neonatal respiratory morbidity12 and adverse infant outcomes such as diarrhea/gastroenteritis,10,11 atopy,11 and asthma.10 The present analyses provide evidence from a Chinese, postindustrialized population that cesarean birth is not associated with hospitalization or outpatient visits during the first 18 months overall. We cannot completely rule out isolated links with afebrile respiratory, gastrointestinal, skin and other minor conditions in the ambulatory setting.
This study has the advantages of longitudinal design, large sample size, high rates of recruitment uptake, and follow-up—thus minimizing selection biases and strengthening generalizability. In addition, we were able to adjust for a range of potential confounders in the multivariable models. Our hierarchical adjustment highlighted the importance of taking into full account the various sets of confounding factors when considering the use of health services as an outcome. Consistent with experience elsewhere, the most important of these was socioeconomic position, with higher SES families having more access to elective medical services. The odds ratios in Table 2 did not change appreciably from the crude estimates until Model 3 incorporated the 2 socioeconomic indicators.
Regarding the few isolated positive associations between a surgical birth and outpatient episodes for minor morbidities (Table 3), we cannot rule out residual confounding by socioeconomic position or other factors, especially in light of the inconsistent timing of the observed relationships across diagnostic subgroups, and the negative overall findings throughout the first 18 months. Type I error could also explain these associations, given the large number of comparisons in the subgroup analyses. If these specific associations are confirmed in other studies, they could perhaps be explained biologically, as has been suggested by some reports in the literature.11,16–20
Several caveats bear mention. First, we were unable to distinguish between primary versus repeat cesarean. Also, we did not have data on the clinical indications or parental choice with respect to method of delivery. However, all analyzed subjects were term singletons, and adjusting for birth order (first born vs. subsequent births) did not markedly change the direction or magnitude of associations. Second, our data relied on parental self-report of method of delivery, frequency and type of health services use, and doctor diagnoses of illness conditions. There is no reason to believe misclassification would be systematic. Third, we relied on medical care utilization rather than underlying illness as outcome measures. This could potentially bias our findings if there were fundamental differences in a family's reliance on doctor or hospital visits to handle infant illnesses that also related to the method of delivery. To compensate, we controlled for highest parental education attainment as well as the type of hospital at birth (public vs. private ownership).
Given the continued rise in cesarean rates, there should be continued surveillance of possible health consequences in both the mother and infant, including follow-up beyond the 18-months period considered here.
We thank all infants and parents in the 1997 Hong Kong Birth Cohort for their participation and the Family Health Service, Department of Health, Government of the Hong Kong Special Administrative Region for collaborating on the study and facilitating recruitment and follow-up.
1. Resnik R. Can a 29% cesarean delivery rate possibly be justified? Obstet Gynecol
2. Cyr RM. Myth of the ideal cesarean section rate: commentary and historic perspective. Am J Obstet Gynecol
3. Dobson R. Caesarean section rate in England and Wales hits 21. BMJ
4. Roberts CL, Tracy S, Peat B. Rates for obstetric intervention among private and public patients in Australia: population based descriptive study. BMJ
5. Belizan JM, Althabe F, Barros FC, et al. Rates and implications of caesarean sections in Latin America: ecological study. BMJ
6. Stanton CK, Holtz SA. Levels and trends in cesarean birth in the developing world. Stud Fam Plann
7. Leung GM, Lam TH, Thach TQ, et al. Rates of cesarean births in Hong Kong: 1987–1999. Birth
8. Leung GM, Lam TH, Wong I, et al. Caesarean section rate in England and Wales. Caesarean section rates in Hong Kong. BMJ
9. Declercq E, Menacker F, MacDorman M. Maternal risk profiles and the primary cesarean rate in the United States, 1991–2002. Am J Public Health
10. Hakansson S, Kallen K. Caesarean section increases the risk of hospital care in childhood for asthma and gastroenteritis. Clin Exp Allergy
11. Laubereau B, Filipiak-Pittroff B, von Berg A, et al. Caesarean section and gastrointestinal symptoms, atopic dermatitis, and sensitisation during the first year of life. Arch Dis Child
12. Many A, Helpman L, Vilnai Y, et al. Neonatal respiratory morbidity after elective cesarean section. J Matern Fetal Neonatal Med
13. Hannah ME, Whyte H, Hannah WJ, et al. Maternal outcomes at 2 years after planned cesarean section versus planned vaginal birth for breech presentation at term: the International Randomized Term Breech Trial. Am J Obstet Gynecol
14. Whyte H, Hannah ME, Saigal S, et al. Outcomes of children at 2 years after planned cesarean birth versus planned vaginal birth for breech presentation at term: the International Randomized Term Breech Trial. Am J Obstet Gynecol
15. Landon MB, Hauth JC, Leveno KJ, et al. Maternal and perinatal outcomes associated with a trial of labor after prior cesarean delivery. N Engl J Med
16. Ahtonen P, Lehtonen O, Kero P. Clostridium perfringens in stool, intrapartum antibiotics and gastrointestinal signs in a neontal intensive care unit. Acta Paediatr
17. Illi S, von Mutius E, Lau S. The pattern of atopic sensitization is associated with the development of asthma in childhood. J Allergy Clin Immunol
18. Malamitsi-Puchner A, Protonotariou E, Boutsikou T, et al. The influence of the mode of delivery on circulating cytokine concentrations in the perinatal period. Early Hum Dev
19. Orrhage K, Nord C. Factors controlling the bacterial colonization of the intestine in breastfed infants. Acta Paediatr
20. Smaill F, Hofmeyr G. Antibiotic prophylaxis for cesarean section. Cochrane Database Syst Rev