Abstract Using a retrospective cohort study design, we report empirical evidence on the effect of parental socioeconomic status, primary care, and health care expenditure associated with preterm or low birth weight (PLBW) babies on their mortality (neonatal, postneonatal, and under-five mortality) under a universal health care system. A total of 4,668 singleton PLBW babies born in Taiwan between Jan 1 and Dec 31, 2001 are extracted from a population-based medical claims database for a follow-up of up to five years. Multivariate survival models suggested the positive effect of higher parental income was significant in neonatal period but diminished in later stages. Consistent inverse relationship was observed between adequate antenatal care and three outcomes: neonatal HR: 0.494, 95% CI: 0.312-0.783; postneonatal HR: 0.282, 95% CI: 0.102-0.774; and under-five HR: 0.575, 95% CI: 0.386-0.857. Primary care services uptake should be actively promoted, particularly in lower income groups, to prevent premature PLBW mortality. Keywords: preterm birth, low birth weight, socioeconomic status, health care disparities, premature mortality Abbreviated Short Title: Socioeconomic status, primary care, and preterm low birth weight survival under social health insurance 1 Introduction From 1971 to 2006, neonatal mortality and infant mortality rates in Taiwan have dropped from 6.72 to 3.01 per 1,000 live births, and 16.56 to 5.35 per 1,000 live births, respectively, while a gradual increase in prevalence of preterm or low-birth-weight (PLBW) births is observed - despite Taiwan’s economic standing that supports a sufficient healthcare capacity.1,2 Taiwan’s sex ratio (SR) at birth, estimated at 1.08 in 2001, also lies above biological norm (only slightly above 1). This pattern is consistent with skewed SRs found in Asian countries where the cultural context favors birth of boys over girls and could suggest sex-based abortion.3,4 Most Taiwanese mothers still live under a traditional norm where son is to be produced to become the family heir and daughter is to be married outside family. However, a study conducted in Taiwan overruled the possibility of selective abortion for the skewed SR and confirmed that the sex imbalance occurs as early as midtrimester.5 Since disclosure of prenatal fetal sex is prohibited in Taiwan, there can be other factors that can attribute to the SR difference such as birth order, parental age, and environmental conditions.6,7 PLBW birth is a major risk factor for infant and child mortality. Preterm births are those that occur at less than 37 complete weeks of gestation. LBW births refer to babies born weighing 2500g or less, regardless of gestational age. McCormick8 found 2 that preterm births account for 75% of perinatal mortality and suggested that neonatal mortality can be effectively diminished by efficacious antenatal services that prevent LBW births. This was confirmed by various subsequent studies.9,10 In Taiwan, prevention of child mortality in the country, however, still falls short of the highest attainable standard in comparison with other industrialized countries.11 This leads to our speculation that lack of primary care (PC) uptake or socioeconomic status may contribute to this pattern. Socioeconomic factors like parental social class can exert influence on the survival of PLBW babies, as a cumulative effect of income on health.12,13 Two hypotheses have been suggested in support of this association: first, children in families with higher income and education have better use of health services as a result of advanced social support and knowledge about health to cushion the impact of health risks14; second, children in socioeconomically-disadvantaged families are at more frequent exposure to injuries and health hazards.15 Low level of household income generates financial barrier for accessing health services for children in need, creating disproportionate accessibility across different socioeconomic classes. Presently, there is a scarcity of literature on the effect of socioeconomic status (SES) and primary care on neonatal, infant and child mortality of PLBW babies for developed countries. In developing countries like Cuba, an universal health care 3 system with emphasis on PC has helped to reduce LBW prevalence which, in turn, lowered the population’s child and infant mortality.16 Inverse effect of PC such as attended births by skill professionals on infant and child mortality has been demonstrated in a study done on 102 low- and middle-income countries.17 Another study examining socially disadvantaged, young unmarried mothers in Malaysia showed that prenatal care visits serve as a risk factor for adverse birth outcomes.18 Even in areas with considerable social gaps, an increased number of PC physicians can help to ameliorate LBW and infant mortality in the United States.19 Nevertheless, whether unfavorable effect on birth outcome exerted by low-SES can be overshadowed by sufficient uptake of PC is under-explored at the individual-level in the scientific literature Unlike previous existing studies, this study investigates Taiwan’s social health insurance that supports a UHC system that has been implemented for almost two decades. Its comprehensive benefit package included an array of available PC services that could be more effectively promoted and utilized, especially for the socioeconomically-disadvantaged. Hence, we hope to look into predictors of PLBW survival with consideration of the UHC system and household socioeconomic factors. Using a retrospective cohort study design, we specifically examine the extent of PLBW health disparities and the effect of parental SES, PC, and health care 4 expenditure associated with PLBW babies on their mortality at different stages of development. Method Data source and study cohort We include PLBW babies born between Jan 1 and Dec 31, 2001 extracted from the National Health Insurance Research Database (NHIRD) using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes 764 and 765, and V21.3. Here we examine the pooled sample of preterm and LBW babies (i.e., PLBW) due to the indistinguishable nature of the ICD-9-CM codes. Maintained by the National Health Research Institutes (NHRI) in Taiwan, the NHIRD is a complete NHI claims-based database accessible to the public upon application which has been used to monitor and assess various policy implementations.20 Unique identification numbers are encrypted for confidentiality but remain consistent throughout all outpatient and inpatient claim files, for us to trace and construct variables by linking the files to obtain information regarding sociodemographics, medical history and diagnoses. Financed through payroll-based premiums and direct government funding, the NHI provides a comprehensive benefit 5 package including ambulatory care, inpatient care, prescription drugs, dental care, and preventive care like antenatal care.21,22 Today, more than 99% of population is enrolled in the single-payer program. Of the 7,913 live PLBW births, 6,038 (76.3%) are selected based on their enrollment in the NHI as dependents of their parents (immediate household members) since their birthday, and their parents’ continuous NHI enrollment throughout the observation period postpartum for ascertainment of their employment status. After removal of multiple births (n=1,370), a total of 4,668 PLBW are identified: employed mothers (n=2,467) and unemployed mothers (n=2,201). A total of 7,907,230 person-days are observed. Babies as dependents of household members other than their parents, and those with unusual birth dates and maternal age of delivery are excluded. Also excluded are stillbirths (ICD-9-CM: 646.0. 651, 652.6, 656.4. 678.1 V27.1-V27.7). The cohort is followed longitudinally for up to five years until Dec 31 2006 or up to the time of ‘event’. This study has been granted ethics approval by the Institutional Review Board at the China Medical University Hospital. Data collection Dependent variables 6 Primary end-points of this study are neonatal mortality (<28 day), postneonatal mortality (28 day–1 year), and under-five mortality (0-5 year). Because mortality rates and their risk factors are remarkably different among these stages, three sub-periods are defined, each carrying out a separate model of analysis. Death or ‘event’ is defined by admission records, either by in-hospital death or discontinuation of insurance status within 30-day of discharge from the hospital, which is proven to be a good proxy for survival.23,24 For these individuals, index date is their date of birth and exit date of insurance enrollment marks the end of their observation period. Loss-of-insurance status is a suitable proxy for mortality in this instance because first, loss of insurance after hospitalization is a reasonable hypothesis when considering a newborn baby born PLBW is admitted right after birth and did not survive after a period of hospitalization. Second, with this proxy, we are able to observe mortality of these PLBW babies across three periods which is consistent with the epidemiological pattern in Taiwan (i.e., highest within 28 days after birth and declines over time). However, any loss-of-insurance status outside these criteria is treated as censored observation. Independent variables 7 We specify four general categories of measures for independent variables. In the sociodemographic category, the parent premium is used for ascertainment of the parental SES and the city of enrollment is used to define the geographical location. It is generally accepted that parental SES is a good and valid indicator for early childhood socioeconomic position. Given the almost 100% coverage of the Taiwan NHI, we believe that the payroll-based premium may well be one of the best available nationwide data that can be used to infer parental income. There are two scenarios under which babies are dependents of parents’ insurance – a) mother as the insurance payer: mother is employed and the baby is mother’s dependent due to relative lower payroll-based premium compared to the father and; b) father as the insurance payer: father is employed and is the head insured person of the household for his spouse and children (i.e., mother most likely does not have a job). Owing to the distinct nature of these types of payroll, we stratify our analyses according to the insurance enrollment status of mothers at time of delivery: hereafter refer to as employed (former) or unemployed (latter). This is performed so that households with same parental employment conditions would be compared. In the PC category, government-subsidized antenatal care is calculated within 37-week before the index date for study cohort. We specify receipt of antenatal care 8 as having at least four antenatal care visits because this number of visits confine to the first two trimesters of gestation which are crucial for prevention of adverse delivery outcome, whereas visits in the third trimester are mainly for detection and treatment of problems whose frequency may vary depending on conditions.8 Similarly, government-subsidized neonatal care and immunization are defined after neonatal period during which major development examinations and vaccinations are received. For health expenditure, we focus on the cost-sharing component of health care financing. Two parameters for health expenditure are derived for analysis: total out-of-pocket (OOP) health spending as a percentage of parental income and total OOP health spending as a percentage of total health expenditure (THE). The former illustrates the proportion of parental income attributed to health care use (ability to pay) and the latter depicts relative financial contribution by individuals from the system standpoint (affordability of care). OOP expenditures are calculated separately for each sub-period by service type. We also consider confounders such as maternal comorbidities (identified within 365.25 days before index date) and physiological conditions originating in the perinatal period (Supplementary Table 1). Data analysis 9 Descriptive analysis Frequency distributions of sociodemographic and physiological attributes are reported for the PLBW birth cohort from 2001, and stratified for employed and unemployed mothers (Table 1). Intra-group comparisons are performed using Student’s t-test, Chi-square test and Fisher’s exact test depending on nature of variables. For period-sensitive predictors, total observation period is reported in person-days (Supplementary Table 2). Health service utilization and expenditure variables are dichotomized into low and high groups using medians of the study cohort. Distribution of resources across income and geographical groups are illustrated in Figure 1. Bivariate analysis We are interested in the relative impact of neighborhood and individual factors on PLBW survival. Effects of relative parental income (individual-level) and neighborhood (community-level) on neonatal mortality are jointly explored in a bivariate Cox proportional hazard regression model stratified by maternal employment status (Figure 2). In exploring the effect of income inequality, we derive relative parental income (parental income divided by median income). 10 Multivariate analysis Survival analysis using Cox proportional hazards model with time-dependent covariates (parental income and geographical area of enrollment) is used to assess influence of parental SES and other risk factors on three outcomes. The assumption of proportionality for each of the models is validated. Two models are developed with sequential inclusion of predictors: Model 1 included parental income and OOP expenditure as share of parental income while controlling for baby sex, geographical area, maternal age and comorbidities, complications at birth, and health service utilization; Model 2 is equivalent to Model 1 with further adjustment for PC use. Health service utilization is adjusted to take into account any additional support or accessibility to care provided by more health conscious parents that would influence PLBW survival rate. Test for interaction does not reveal higher service utilization (outpatient and inpatient) is associated with greater OOP expenditure, suggesting that inclusion of interaction terms in the final multivariate model is not required (Model 2 in Table 2). Analyses are performed with exclusion of multiple births with consideration of the biased effect caused by immediate physiological factors not related to SES, but results inclusive of multiple births are also provided (Supplementary Tables 3-4). Multicollinearity tests are 11 conducted to ensure no violation of models is evident. For all statistical analyses, P-value of less than 0.05 is considered to be statistically significant. All analyses are performed using SAS statistical package version 9.2 (SAS Institute, Cary, NC). Results Descriptive analysis Table 1 shows the baseline characteristics of the singleton children born PLBW. Distribution of parental income at birth is different when comparing PLBW babies between employed and unemployed mothers, where the latter group has a higher mean income (NTD26050.8 vs. NTD24633.1; USD1≈NTD33.8 in 2001) – a difference of roughly 6%. For time-sensitive variables, very high inpatient OOP expenditure as share of parental income was spent in the neonatal period and declined as time progressed (Supplementary Table 2). A slightly higher proportion of richest 20% (Q5) mothers received adequate prenatal care compared with other income groups while the poorest mothers spent the highest proportion of income on health within 28 days after birth (Figure 1). In general, excess mortality rates are highest among neonatal PLBW babies born in the 12 bottom two income quintiles or in the south and east regions of the country (data not shown). Bivariate and multivariate survival analysis When jointly considered the effect of individual- and community-level factors on neonatal mortality in bivariate analysis, singleton PLBW babies born to employed mothers with relative low income residing in the south region are at highest risk (hazard ratio (HR): 3.623; P<0.001) (Figure 2). In Model 1 of the multivariate analysis (Table 2), relative high income is associated with lower mortality risk consistently through neonatal, postneonatal, and under-five year periods in employed mothers (HR: 0.528, 95% CI: 0.326, 0.857, P<0.01; HR: 0.295, 95% CI: 0.113, 0.769, P<0.05; HR: 0.505, 95% CI: 0.334, 0.765, P<0.01, respectively). High outpatient and inpatient OOP as share of parental income is also consistently associated with lower mortality risks for the three periods as compared to relative low group. Further adjustment of PC use in Model 2 diminished the effect of relative high parental income in employed mothers, but suggested a profound inverse relationship between adequate antenatal care and neonatal mortality, postneonatal mortality, and under-five mortality (HR: 0.494, 95% CI: 0.312, 0.783, P<0.01; HR: 0.282, 95% CI: 13 0.102, 0.774, P<0.05; and HR: 0.575, 95% CI: 0.386, 0.857, P<0.01, respectively). Furthermore, high neonatal care and high subsidized immunization are significant protective agents for postneonatal and under-five mortality in both employment groups, respectively. For OOP expenditure (inpatient and outpatient), its inverse effect appear to persist from neonatal period up to five years of age in PLBW children of both groups (i.e., an employed or unemployed mother with relative high OOP expenditure as a proportion of parental income will have a lower likelihood of having a singleton PLBW baby experiencing premature death compared to relative low OOP spending counterpart). Discussion Main findings of this study Our objective is to investigate whether after the implementation of universal health care system in Taiwan, distribution of PLBW survival will be equitable in the presence of an upstream factor like social position by income when controlling for PC uptake and OOP health expenditure. We demonstrated that in a health system financed partly by direct payments at point of service, inequalities in service utilization and OOP expenditure up to five years of age still exist for PLBW babies across SES groups 14 and geographical regions. This implies uneven distribution of resources amongst our population in spite of the universal coverage insurance. Access to PC, for instance, is enjoyed most by the better-off while OOP payments seem to generate the most financial burden in poorer households. More importantly, PC services are positively associated with PLBW survival up to five years after birth, while high relative parental income exerts an impact only in the first 28-days. What is already known on this topic SES is an important input in the child health production. Effect of parental income on children’s health has been well established.12-15,25-27 SES can mitigate the consequences of health shocks, and higher SES can augment the chance of children recovering from negative health conditions.26 Cost-sharing, alternatively, reduces health care accessibility more among the poor than the rich. Distributions of long-term medication use and health resource use are found to be more concentrated amongst higher social classes,25,28 and disparities in physician visits across different SES classes are considerable in industrialized countries with heavy reliance on private financing like patient cost-sharing.29 So we can argue if SES were to exert an effect on PLBW survival, it would be that the high-SES mothers had higher access to helpful resources during gestation and early intervention programs for 15 PLBW babies than low-SES mothers.15,30 Thus, health status of low-SES children could eventually catch up to that of high-SES children. This conclusion indeed supports our finding that the effect of relative parental income is significant in the neonatal period but dissipates thereafter for PLBW children. Studies have also highlighted the importance of PC provision in enhancing children’s health and avoiding premature deaths in both developed and less developed economies.31-34 Health systems that center around PC have better health outcomes.32,33 The role of PC is multifaceted. It not only aids in early prevention of disease, but also increases access to health services for relatively disadvantaged groups. Our results consent with this argument when we observe the negative association between adequate uptake of antenatal care, neonatal care, immunization and PLBW mortality even controlling for individual-level SES. However, we also agree that gaps in the receipt of ample PC services persist in many areas.35 Limitations of this study We observed a high SR of 1.23 for our PLBW cohort. The reason for this could be the cultural-based sex-selectivity that is amplified in PLBW babies, i.e., more sex-based abortions occur for female PLBW babies than male PLBW babies given the adverse birth outcome. Although prenatal sex-screening is banned in Taiwan, it may 16 not have been well enforced in 2001 and the determination of sex would affect parents’ choice for abortion. Since female PLBW babies are more likely fall out of our observation and not being included in our selection, our observed mortality could be underestimated. Another caveat from our selection criteria is that we did not include 1,875 (23.7%) babies for reasons of trying to derive an accurate parental income measure: we excluded 1) babies who are not NHI dependents of their father or mother (we cannot infer parental income); or 2) babies who are NHI dependents of their father or mother but their enrollment did not persist throughout the observation period (to avoid unusual cases of frequent, irregular exit-and-re-enrollment activity within the 5-year observation and can greatly fluctuate the observed SES). Smoking in mothers could not be identified. Instead, we controlled for the effect of chronic lung diseases which is a good indicator for chronic smokers.36 Because Taiwan holds low infant mortality and under-five mortality, data from bigger cohort collected from multiple years instead of one year could be more favorable, but this would restrict the length of observation period due to data availability which we did not wish to forgo. What this study adds 17 In spite of these caveats, we find only few published research investigating socioeconomic disparities in PLBW child mortality with relation to system-level and individual-level predictors. Our findings support the universal coverage of essential PC services for pregnant women, babies and infants irrespective of individual SES in order to improve PLBW survival. Furthermore, we measured long-term impacts of relative poverty. We derived parental SES using salary-based premium paid by each enrollee which is a reliable measure of material living standards. We further considered relative income a measure of social exclusion since it was proposed that wellbeing is more closely related to relative income than absolute income in rich countries.37 The large sample size and adequate length of follow-up time allow for an invaluable opportunity to examine predictors of the postnatal survival of a PLBW birth cohort. Interestingly, we observe that PLBW babies who incurred high inpatient and outpatient OOP payments as a proportion of their parental income resulted in better odds of survival. This indicates that the survival of a less physically-sound PLBW baby depends not so much on how large their parents' financial sacrifice is, but on how much money is spent on the baby, independent of parental income. This is not in alignment with what literature has been suggesting about catastrophic health spending leading up to poorer and inequitable health.38,39 We would like to interpret 18 this as these PLBW babies are at greatly heightened risk of life-threatening illness at baseline comparing with general population. Hence, we see that ‘paying for health care’ is beneficial for their survival at least before reaching their fifth birthday. In the treatment-oriented Taiwanese society, it is possible that parents would seek extra care even if it exceeds their health spending for the cost of saving their PLBW child. Conclusion In concluding, we find that social gradient in health service utilization exists in Taiwan’s social health insurance system and that sufficient PC uptake can positively associate with survival of high-mortality-risk PLBW babies up to five years after birth, independent of parental income. This finding confirms that even under universal health care scheme, the use of PC services should be actively promoted especially in the lower SES groups. 19 Acknowledgements This study is based on data provided and managed by the Taiwan National Health Research Institutes (NHRI). We thank the NHRI for providing NHIRD and administrative support for research purposes (Project No. 98348). Declaration of Conflicting Interest The authors declare there is no conflict of interest. Funding This work was supported in part by the National Science Council, Executive Yuan, Taiwan (NSC 101-2314-B-039-016). 20 References 1. 4. NHRI (National Health Research Institutes). Healthy People 2020 Technical Report. Miaoli: National Health Research Institutes, Taiwan; 2008. BHP (Bureau of Health Promotion). Bureau of Health Promotion 2011 Annual Report. Taipei: Department of Health, Executive Yuan, Taiwan; 2011. Jha P, Kesler MA, Kumar R, Ram F, Ram U, Aleksandrowicz L, Bassani DG, Chandra S, Banthia JK. Trends in selective abortions of girls in India: analysis of nationally representative birth histories from 1990 to 2005 and census data from 1991 to 2011. Lancet. 2011;377:1921-1928. Zhu WX, Lu L, Hesketh T. China’s excess males, sex selective abortion, and one 5. child policy: analysis of data from 2005 national intercensus survey. BMJ. 2009;338:b1211. Lee IW, Lai YC, Kuo PL, Chang CM. Human sex ratio at amniocentesis and at birth 2. 3. 6. 7. 8. in Taiwan. Taiwan J Obstet Gynecol. 2012;51:572-575. James WH, Rostron J . Parental age, parity, and sex ratio in births in England and Wales, 1968–1977. J Biosoc Sci. 1985;17:47–56. Smith D, Von Behren J. Trends in the sex ratio of California births, 1960-1996. J Epidemiol Community Health. 2005;59:1047-1053. McCormick MC. The contribution of low birth weight to infant mortality and childhood morbidity. N Engl J Med. 1985;312(2):82-90. 9. Wilcox AJ, Skjaerven R. Birth weight and perinatal mortality: the effect of gestational age. Am J Public Health. 1992;82(3):378-382. 10. Kramer MS, Demissie K, Yang H, Platt RW, Sauvé R, Liston R. The contribution of mild and moderate preterm birth to infant mortality. Fetal and Infant Health Study Group of the Canadian Perinatal Surveillance System. JAMA. 2000;283(7):843-849. 11. Wu JCL, Chiang TL. Comparing child mortality in Taiwan and selected industrialized countries. J Formos Med Asso. 2007;106:177-180. 12. Barker DJP. Mothers, babies, and disease in later life. London: BMJ Publishing Group; 1994. 13. Marmot M, Wadsworth MEJ, eds. Fetal and early childhood environment: long-term health implications. British Medical Bulletin. 1997;53 (Issue 1). 14. Propper C, Rigg J, Burgess S. Child health: evidence on the roles of family income and maternal mental health from a UK birth cohort. Health Econ. 2007;16:1245-1269. 15. Currie J, Stabile M. Socioeconomic status and child health: Why is the relationship stronger for older children? American Economic Review. 21 2003;93:1813-1823. 16. Lopez NB, Choonara I. Can we reduce the number of low-birth-weight babies? The Cuban experience. Neonatology. 2008 [Epub ahead of print] DOI: 10.1159/000155649. 17. Hsieh VC, Wu JC, Wu TN, Chiang TL. Universal coverage for primary health care is a wise investment: evidence from 102 low- and middle-income countries. Asia Pac J Public Health. 2013 [Epub ahead of print]: DOI:10.1177/1010539513492562. 18. Mohd Zain N, Low WY, Othman S. Impact of maternal marital status on birth outcomes among young Malaysian women: a prospective cohort study. Asia Pac J Public Health. 2014 [Epub ahead of print]: PII:1010539514537678. 19. Shi L, Macinko J, Starfield B, Xu J, Regan J, Politzer R, Wulu J. Primary care, infant mortality, and low birth weight in the States of the USA. J Epidemiol Community Health. 2004;58:374-380. 20. Wu TY, Majeed A, Kuo KN. An overview of the healthcare system in Taiwan. London Journal of Primary Care. 2010;3:115-119. 21. Chiang TL. Taiwan’s 1995 health care reform. Health Policy. 1997;39:225-239. 22. Lu RJF, Chiang TL. Evolution of Taiwan’s health care system. Health Econ Policy Law. 2011;6:85-107. 23. Lien HM, Chou SY, Liu JT. Hospital ownership and performance: evidence from stroke and cardiac treatment in Taiwan. J Health Econ. 2008;27:1208-1223. 24. Shen HN, Lu CL. Incidence, resource use, and outcome of acute pancreatitis with/without intensive care: a nationwide population-based study in Taiwan. Pancreas. 2011;40:10-15. 25. van Doorslaer E, Koolman X, Jones A. Explaining income-related inequalities in doctor utilization in Europe. Health Econ. 2004;13:629-647. 26. Kruk KE. Parental income and the dynamics of health inequality in early childhood – evidence from the UK. Health Econ. 2013;22:1199-1214. 27. Pickett KE, Wilkinson RG. Child wellbeing and income inequality in rich societies: ecological cross sectional study. BMJ. 2007;335:1080. 28. Gissler M, Rahkonen O, Järvelin MR, Hemminki E. Social class differences in health until the age of seven years among the Finnish 1987 birth cohort. Soc Sci Med. 1998;46:1543-1552. 29. Allin S, Masseria C, Mossialos E. Measuring socioeconomic differences in use of health care services by wealth versus by income. Am J Public Health. 2009;99:1849-1855. 30. Currie J, Hyson R. Is the impact of health shocks cushioned by socioeconomic status? The case of low birthweight. American Economic Review. 22 1999;89:245-250. 31. Guttmann A, Shipman SA, Lam K, Goodman DC, Stukel TA. Primary care physician supply and children’s health care use, access, and outcomes: findings from Canada. Pediatrics. 2010;125:1119-1126. 32. Bhutta ZA, Ali S, Cousens S, et al. Alma-Ata: Rebirth and revision 6 – Interventions to address maternal, newborn, and child survival: what difference can integrated primary health care strategies make? Lancet. 2008;372:972-989. 33. Starfield B, Shi L, Macinko J. Contribution of primary care to health systems and health. Milbank Q. 2005;83:457-502. 34. Jones G, Steketee RW, Black RE, Bhutta ZA, Morris SS. How Many Child Deaths Can We Prevent This Year? Lancet. 2003;362:65–71. 35. Stevens GD, Shi L. Racial and Ethnic Disparities in the Quality of Primary Care for Children. Journal of Family Practice. 2002;51:573. 36. Barnes PJ, Shapiro SD, Pauwels RA. Chronic obstructive pulmonary disease: molecular and cellular mechanisms. Eur Respir J. 2003;22:672-688. 37. Marmot M, Wilkinson RG. Psychosocial and material pathways in the relation between income and health: a response to Lynch et al. BMJ. 2001;322:1233-1236. 38. CSDH (Commission on Social Determinants of Health). Closing the gap in a generation: health equity through action on the social determinants of health. Final Report of the Commission on Social Determinants of Health. Geneva: World Health Organization; 2008. 39. Yates R. Universal health care and the removal of user fees. Lancet. 2009;373:2078-2081. 23 Table 1. Descriptive profile of sociodemographic and physiological attributes for study cohort at baseline in 2001. . Characteristic All PLBW Employed mothers Unemployed mothers n=4668 n=2467 n=2201 Person-days 7,907,230 4,211,841 3,695,389 Sociodemographics Sex, no. (%) Female 2093 (44.84) 1109 (44.95) 984 (44.71) Male 2575 (55.16) 1358 (55.05) 1217 (55.29) Geographical area, no. (%) ** North 2273 (48.69) 1276 (51.72) 997 (45.30) Centre 1065 (22.81) 502 (20.35) 563 (25.58) South 1121 (24.01) 574 (23.27) 547 (24.85) East 175 (3.75) 101 (4.09) 74 (3.36) Islands and others 34 (0.73) 14 (0.57) 20 (0.91) Maternal age at delivery, no. (%), years ** <20 159 (3.41) 62 (2.51) 97 (4.41) 20-29 2357 (50.49) 1097 (44.47) 1260 (57.25) 30-39 2015 (43.17) 1220 (49.45) 795 (36.12) ≧40 137 (2.93) 88 (3.57) 49 (2.23) Parental income at birth, no. (%), NTD ** Q5 (≦16500) 919 (19.69) 414 (16.78) 505 (22.94) Q4 (16501-19200) 834 (17.87) 502 (20.35) 332 (15.08) Q3 (19201-25200) 667 (14.29) 417 (16.90) 250 (11.36) Q2 (25201-38200) 1231 (26.37) 631 (25.58) 600 (27.26) Q1 (>38200) 1017 (21.79) 503 (20.39) 514 (23.35) Parental income at birth, mean±SD (NTD) 25301.6±14737.4 24633.1±12885.7 26050.8±16539.7 * Physiological attributes Maternal comorbidities, no. (%) Diabetes mellitus 280 (6.00) 197 (7.99) 83 (3.77) ** Chronic renal disease 6 (0.13) 5 (0.20) 1 (0.05) Hypertension 325 (6.96) 234 (9.49) 91 (4.13) ** Chronic lung disease 82 (1.76) 55 (2.23) 27 (1.23) * Complications at birth, no. (%) Congenital anomalies 1402 (30.03) 739 (29.96) 663 (30.12) Birth trauma 95 (2.04) 45 (1.82) 50 (2.27) RDS 1510 (32.35) 785 (31.82) 725 (32.94) Note: Values are n (%) or mean ± SD. *P<0.05; **P<0.01 PLBW - preterm low birth weight; NTD - New Taiwan Dollar; Q - quintile; RDS - respiratory distress syndrome 24 Table 2. Period-specific hazard ratios using Cox Proportional Hazards Model for the PLBW mortality at 0-28 day, 28 day-1 year, and 0-5 year observation periods in employed and unemployed mothers. Employed Mothers Variable† Model 1 HR (95% CI) Model 2 HR (95% CI) Unemployed Mothers Model 1 HR (95% CI) Model 2 HR (95% CI) <28d 0.728 (0.476-1.116) 0.743 (0.486-1.135) Relative income 0.528 (0.326-0.857)** 0.569 (0.347-0.931)* 0.766 (0.472-1.242) Antenatal care (at least 4 times) -0.494 (0.312-0.783)** -OOP Outpatient (as proportion of parental income) 0.000 (0.000-) 0.000 (0.000-) 0.000 (0.000-) 0.000 (0.000-) 0.661 (0.420-1.042) 0.691 (0.439-1.087) OOP Inpatient (as proportion of parental income) 0.473 (0.284-0.787)** 0.479 (0.288-0.798)** 28d-1yr Relative income 0.295 (0.113-0.769)* 0.485 (0.163-1.442) 0.815 (0.369-1.802) 0.703 (0.298-1.661) Antenatal care (at least 4 times) -0.282 (0.102-0.774)* -0.323 (0.074-1.403) Neonatal care -0.037 (0.005-0.285)** -0.058 (0.013-0.253)** Subsidized immunization -0.067 (0.009-0.521)** -0.000 (0.000-) OOP Outpatient (as proportion of parental income) 0.049 (0.006-0.382)** 0.051 (0.006-0.407)** 0.163 (0.048-0.560)** 0.118 (0.031-0.444)** OOP Inpatient (as proportion of parental income) 0.077 (0.017-0.345)** 0.078 (0.017-0.354)** 0.285 (0.119-0.684)** 0.280 (0.113-0.693)** 0-5y Relative income 0.505 (0.334-0.765)** 0.788 (0.520-1.194) 0.551 (0.388-0.780)** 0.767 (0.541-1.089) Antenatal care (at least 4 times) -0.575 (0.386-0.857)** -0.723 (0.462-1.131) Neonatal care -0.000 (0.000-) -0.000 (0.000-) Subsidized immunization -0.167 (0.067-0.416)** -0.075 (0.024-0.237)** OOP Outpatient (as proportion of parental income) 0.018 (0.003-0.134)** 0.021 (0.003-0.154)** 0.031 (0.008-0.128)** 0.051 (0.013-0.207)** OOP Inpatient (as proportion of parental income) 0.453 (0.292-0.703)** 0.581 (0.375-0.900)* 0.412 (0.272-0.622)** 0.482 (0.318-0.730)** Model 1: adjusted for baby sex, geographical area, maternal age and maternal comorbidities, complications at birth, and health service utilization (outpatient frequency and inpatient length) Model 2: Model 1 + primary care use (antenatal care, neonatal care, subsidized immunization) †: All variables dichotomized by median (high vs. low, yes vs. no) *P<0.05; **P<0.01 25