The study intervention population included a consecutive series of in-hospital live-born infants with any form of isolated or associated typical CL/P, between January 1, 2003 and December 31, 2005 in 47 hospitals in Argentina, Brazil, Bolivia, Chile, Colombia, Ecuador and Venezuela [26]. The isolated group had no other major birth defects besides CL/P and no syndromes including those with or without chromosomal abnormalities. Pierre-Robin was considered as isolated cleft palate. Associated cases were identified using physical exams and medical expert review [27]. The control population included all infants with typical CL/P born between January 1, 2001 and December 31, 2002 in the same study hospitals as the intervention group. The same definitions for isolated and associated status were used for the treatment and control groups. Table 1 reports the distribution of the other birth defects in the associated treatment and control groups. The investigators considered randomizing the prospectively enrolled group into control and intervention groups to be unethical given that the study's intervention represents the standard of care in developed countries such as the United States, where early and regular evaluation of infants with oral clefts within the neonatal period is emphasized [28].

The study hospitals were affiliated with the Latin American Collaborative Study of Congenital Malformations (ECLAMC), which is a birth defect surveillance and research program throughout South America [29]. ECLAMC involves collaboration with pediatricians who monitor birth defect occurrence in their hospitals and collect birth record data on affected and unaffected infants. The ECLAMC pediatricians were responsible for identifying and enrolling infants into the study and providing the intervention.

The intervention sample included 663 infants enrolled within 48 hours after birth and before hospital discharge. The sample represents 94% of eligible potential subjects (707 cases). The other potential subjects did not consent to the study. Parental consent was obtained from each child's parent(s) before participation.

The control sample included 456 infants who represent 91% of potential control subjects (499 cases). The non-intervened control group had received the medical care available in the community, which is unmeasured but is thought to vary between the various communities included in the study and to be less intensive and systematic on average than the study intervention. The study pediatricians identified the control group from the ECLAMC and hospital birth records and conducted phone and/or in-person interviews with parents to obtain data on infant survival and hospitalizations. The pediatricians also abstracted related information from the infants' hospital charts.

As described below, the intervention started at the 7^th day of life. Therefore, all infants who died before the 7^th day were excluded from the analysis as they had not received the intervention as defined in the protocol, resulting in a final sample of 543 infants in the intervention group and 372 controls for the survival analysis.

Table 2 presents a comparison of baseline health and other relevant characteristics between the study populations and enrolled samples by intervention status. There were no statistically significant differences between the study populations and samples, suggesting no sample selection bias due to parental consenting into the study. The sample is geographically and socioeconomically diverse, which enhances its representativeness of the infant populations in the study countries. The majority of infants in the study countries are thought to be born in healthcare institutions [30]. The WHO reports that rates of births attended by skilled health personnel, most of which are expected to have occurred in healthcare institutions, ranged from about 61% in Bolivia to ~100% in Chile in 2003-2005 [31]. The study results are generalizable to in-hospital born infants, who represent the majority of infants in the study populations.

The study was approved by the University of Iowa IRB (protocol ID 200109029).

The intervention consisted of four weekly visits with the study pediatricians between the 7^th and 28^th day of life which involved systematic evaluation of the infant's health, providing pediatric care, referral to specialized care as needed and educating parents in feeding practices and optimal infant care and monitoring. First, the pediatricians screened infants for eligibility and enrolled those who are eligible into the study within the first 48 hours and before hospital discharge. As part of this screening, the pediatricians advised mothers about needed healthcare and optimal feeding and homecare practices following ECLAMC'S routine ORIENT program, which provides affected families with cleft- related information and orients them to available healthcare resources [29]. The control families had also received the ORIENT program information at this period.

Once a week, the study pediatricians monitored and assessed the health of the intervention group infants during in-person visits to the pediatricians' practices. The pediatricians conducted home visits when possible with families who missed the weekly visits. The pediatricians evaluated changes in weight, length, head circumference, feeding methods and medical complications such as hyperbilirubinemia or infection. After the examination, the pediatricians referred the infants for further intervention to specialized professionals as needed. The pediatricians assessed the need for hospitalizing infants who showed no weight gain between visits and for more frequent follow-up than the weekly visit schedule for these infants or infants with other medical complications. The study protocol required the pediatricians to hospitalize infants who show more than ten percent weight decrease between visits and to evaluate the need for hospitalizing infants with ten percent or less weight decrease based on the infant's health, medical complications and standard pediatric judgment. The study protocol required the pediatricians to re-emphasize at each visit the recommendations and instructions to parents on feeding practices and optimal homecare. The control group did not receive the weekly pediatric visits provided to the intervention group.

The study assessed the intervention effect on neonatal mortality between the 7^th and 28^th day of life and on hospitalization days during this period among surviving infants, both pooling and stratifying by CL/P associated status. These outcomes were measured beginning on the 7^th day of life given that the pediatric intervention started on that day. We also report the overall neonatal mortality rates between the first and 28^th day of life given that such estimates are rare in South America.

The study pediatricians assessed survival and hospitalizations using the same study questionnaires at all hospitals during their visits with the intervention group and interviews with the control group. Further details of data collection and management are described elsewhere [26].

Given the non-randomized design of the study, we assessed the intervention effects on the study outcomes adjusting for relevant baseline characteristics including infant's cleft type (cleft lip, cleft palate, cleft lip with palate), birth weight, gestational age, sex and ethnic ancestry, maternal age, number of prenatal visits, maternal chronic illness, multiple birth status, cesarean-delivery and prenatal CL/P diagnosis. The regression models included country fixed effects in order to account for country differences. As a reference, we also estimated regressions that did not adjust for baseline characteristics but only included the treatment and country indicators. ECLAMC pediatricians routinely measure these baseline characteristics through interviews with the mothers after delivery and before hospital discharge and through hospital record abstraction. Table 3 describes the baseline characteristics in the study intervention and control groups. There were no significant differences in these characteristics between the intervention and control groups except for a greater number of prenatal visits and greater birth weight among the isolated control group, and a higher rate of prenatal CL/P diagnosis among the isolated intervention group.

The mortality outcome model was estimated by logistic regression. The hospitalization day model was estimated by zero-inflated negative binomial regression in order to account for the large percentage of zero hospitalizations. The variance-covariance matrices were estimated with a Huber-type estimator that is robust for within-hospital clustering [32].

The intervention effects were estimated pooling and stratifying by CL/P associated status given the expected effect differences by the presence of other birth defects. The models for total and associated samples included the number of birth defects other than CL/P as a covariate in order to account for the severity of associated status.

About 6.8% of the sample were lost to follow-up and had no observable mortality outcomes. Table 2 presents a comparison between the enrolled sample and the sample with known mortality status. There were no significant differences in baseline characteristics between these samples for both the intervention and control groups, suggesting no sample selection bias. Therefore, we did not adjust the mortality and hospitalization models for sample selection.

Table 4 reports the mortality rates between the 7^th and 28^th day in the intervention and control groups and the odds ratios (ORs) for the intervention effect from the logistic regression. Table 5 reports the ORs for all model variables.

There were no significant differences in the 7^th-28^th day mortality rates between the intervention and control groups, both pooled and stratified by associated status. In the total sample, the mortality rate was 6.1% in the intervention group versus 4.8% in the control group. In the isolated group, the mortality rates were 2% versus 1.3% in the intervention versus control groups, respectively. In the associated group, the mortality rates were 16.8% versus 20.9% in the intervention and control groups, respectively.

The overall neonatal mortality rates between the first and 28^th day of life were 15.3% and 19.7% in the intervention and control groups, respectively. The overall neonatal mortality rates were 4.7% and 3.8% in the isolated intervention and control groups, respectively, and 37.1% and 58.6% in the associated intervention and control groups, respectively.

Table 6 reports hospitalization days and the intervention effects on hospitalization days between the 7^th and 28^th days of life among the infants who survived the neonatal period. Table 7 reports the effects of all model variables.

The intervention group had fewer neonatal hospitalization days on average compared to the control group (1.9 versus 3.1 days). In the multivariate model, the intervention reduced hospitalization days by about 0.6 days in the total sample (p = 0.089). There was no significant difference in hospitalization days between the isolated intervention and control groups (1.6 versus 1.9 days, respectively). In contrast, there were significantly fewer hospitalization days in the intervention group with associated CL/P compared to the control group (3.4 versus 10.3 days, respectively). In the multivariate model, the intervention reduced hospitalization days in the associated group by about 6 days (p < 0.001).

Table 8 reports the distribution of reasons for hospital readmission for the treatment group; data is not available on hospital readmission reasons for the control group. About 18% of readmissions were due to respiratory problems and 16% were due to weight loss problems. About 12% were due to feeding or nutrition problems. Another 10% were due to surgery including cleft repair or any other surgery, and another 10% were due to icterus syndromes.

We describe below the significant effects of main baseline infant and maternal characteristics on mortality and hospitalization days between the 7^th and 28^th day of life (see detailed results in Tables 5 and 7). African and native infant ancestry, as reported by the mother, increased mortality risk by about 32 and 22 times in the total sample, respectively, compared to other ancestries. Prenatal CL/P diagnosis decreased mortality risk by about 0.1 times in the isolated group (marginally significant). Higher birth weight reduced mortality significantly both pooling and stratifying by associated status - a 300-gram birth weight decreased mortality risk by about 0.7 times in the total sample. In the isolated group, infants with cleft lip only and cleft lip with palate had higher mortality risks by about 7.6 and 6.7 times, respectively, compared to infants with cleft palate only. Finally, the number of other malformations besides CL/P increased mortality risk in the associated group by 1.5 times per malformation.

Several factors had significant effects on hospitalization days in the total sample, primarily due to effects on the isolated CL/P group. Birth weight and gestational age each reduced hospitalization days by about one day per 1000 grams and five gestational weeks, respectively. Infants with cleft lip only and with cleft lip with palate in the total sample had fewer hospitalization days by about 1.5 and 0.6 days, respectively, compared to infants with cleft palate only. Also, infants born through cesarean delivery had longer hospitalization by about one day. Similar effects were seen on hospitalization in the isolated group, in addition to a small decrease in hospitalizations for cases diagnosed prenatally (half a day on average; marginally significant) and an increase in hospitalization days among those whose mothers had chronic illnesses during pregnancy. (0.7 days on average).