Neonatal Tetanus Mortality in Veracruz, Mexico, 1989l WCTOR M. CARDENAS AYALA,~ ROSAMAR~A NXZJAEZURQUIZA,~ DONNA R. BROGAN,~JORGEM. IBARRA ROSALES,~ NOB GATICA VALDBS,~ TERRENCEE. SMITH,~ SALVADOR GALVAN ARRIAGA,~ MAMMAESTRELLA FLORES COLLINS,~ & ALEJANDRO ESCOBARMESAS This article describes a survey conducted in the State of Veracruz, Mexico, to estimate neonatal tetanus (NNT) mortality. The survey, which entailed visits to 72 720 households, collected data on 8 401 live births and 209 infant deaths occurring between April 1988 and May 1989. Twenty-six of the 209 fatalities conformed to a WHO standard case definition of death from neonatal tetanus. The estimated neonatal tetanus mortality was thus 3.1 deaths per 1 000 live births (95% confidence limits = 1.7, 4.5). Comparison of this rate to reported figures suggests that for every NNT death recorded in Veracruz during the study period, as many as 50 others went unreported. A case-control study nested within the survey was conducted to assess preventable NNT risk factors. Limited information on 13 NNT deaths and 217 controls showed an increased risk for neonates who were delivered at home and whose parents’ ethnic background was Mexican Indian. Five of the 13 fatalities had their umbilical cords cut with a domestic or traditional cutting tool such as a reed cane, as compared to none of the 217 controls. The observed vaccine efficacy of 2 + doses of tetanus toxoid was 70% (95% confidence limits = 52,100). Both the mothers of neonates who died of NNT and their controls missed an average of five opportunities to receive tetanus toxoid. These findings underscore the need to launch a perinatal health program serving Mexico’s high-risk populations. eonatal tetanus (NNT), a target dis- zation (EN), is perhaps the most refractory N ease of the World Health Organ- of those diseases to current strategies. ization’s Expanded Program on Immuni- This is partly because in many places im- iThis article will also be published in Spanish in Dr. Manuel Arenas Loyo, Health District 11, the Boletfn de Iu Oficina Sanitaria Panamericana, Vol. Coatzacoalcos; Dr. Rosa Maria Espejo and Dr. Ale- 119, 1995. jandro Escobar Mesa of the Department of Epi- 2The affiliations of the coauthors and other mem- demiology, Xalapa; Dr. Maria Estrella Flores Col- bers of the Veracruz Neonatal Tetanus Mortality lins, Dr. Rosa Maria Nuiiez Urquiza, Dr. Jorge M. Survey Group: Dr. Gabriel Fernandez Rodriguez, Ibarra Rosales, Prof. Salvador Galvan Arriaga, Dr. Veracruz State Health Secretariat, Health District Laura Tirado, and Dr. Victor M. Cardenas Ayala 1, Panuco; Dr. Providencio Martinez Palestino, of the National Institute of Public Health, Cuer- Health District 2, Tuxpan; Dr. Armando Bustos navaca; Dr. No6 Gatica ValdCs, Dr. Lourdes Ca- Rivera, Health District 3, Poza Rica; Dr. Saturnino macho, Dr. Guadalupe Quiroz, Dr. Mauricio Her- Navarro Ramfrez, Health District 4, Martinez de la nandez Avila, and Dr. Cuauhtemoc Ruiz Matus of Torre; Dr. Julio Arronte Arguelles, Health District the Division of Epidemiology, Health Secretariat, 5, Xalapa; Dr. Humberto Vidal Cervantes, Health Mexico City; Dr. Terrence E. Smith of the Univer- District 6, C6rdoba; Dr. Marco A. Rodriguez Vivas, sity of California School of Public Health at Berke- Health District 7, Orizaba; Dr. Gloria Mendoza Ruiz, ley, U.S.A.; and Dr. Donna R. Brogan, Director of Health District 8, Veracruz; Dr. Luis Fernando An- the Division of Biostatistics of the School of Public tiga, Health District 9, Cosamaloapan; Dr. Evaristo Health of Emory University, Atlanta, Georgia, Gonzalez, Health District 10, San And& Tuxtla; U.S.A. 116 Bulletin of PAHO 29(2), 1995 munizers must target two groups: young The 1989 study reported here was car- girls (to protect their future offspring) and ried out in Mexico’s large (71 699 km2) and women of reproductive age (the main tar- populous (6 832 271 inhabitants) State of get group in countries without long- Veracruz. Two previously published re- standing immunization programs). ports (II, 12) had documented a very high Following NNT declines in the devel- incidence of NNT there, especially in rural oped world and paralleling reported de- areas. Registered tetanus mortality among clines in most developing countries (l), Veracruz infants in 1956-1960 was 1.5 deaths since the 1930s Mexico has seen a sharp per 1 000 live births; and while only seven decline in recorded NNT deaths. The fig- cases were reported to the Veracruz Health ures have gone from 8 200 recorded in- Secretariat in 1987, some 35 deaths caused fant deaths from NNT (10 per 1 000 live by NNT were officially registered that year. births) in 1930 to 4 053 (2 per 1 000) in These deaths made the state’s reported 1957 (2) to an average of about 300 an- NNT mortality (approximately 0.15 deaths nually (0.2 per 1 000) in the early 1980s per 1 000 live births) one of the highest in (3). However, the accuracy of these re- the country. corded data is open to question, partly Our study had the following principal aims: because Mexico’s national disease sur- (1) to estimate the number of NNT deaths veillance system has reported fewer NNT in Veracruz between 3 April 1988 and 15 cases than the number of NNT deaths May 1989, together with the magnitude of recorded on death certificates (3). suspected underreporting and (2) to assess In 1988 and 1989 the respective NNT preventable risk factors for NNT. incidences reported for Mexico were 0.053 and 0.043 cases per 1000 live births (4). METHODS According to 1988 National Health Survey figures (5), the percentage of newborns de- Probability Sample Survey livered at hospitals ranged from 90% in Northern Mexico to just under 50% in Cen- The inference population for the sam- tral and Southern Mexico. A 1990 national ple survey was all of the babies born in survey estimated coverage with at least two the State of Veracruz between April 1988 doses of tetanus toxoid to be 32.3% among and May 1989. These babies were iden- women of reproductive age (5). tified by selecting and visiting a proba- Underestimation of this public health bility sample of housing units. In turn, problem by routine surveillance can lead the housing units were selected by means to allocation of fewer resources for im- of a three-stage cluster process. munization. Within this context, special Using systematic random sampling, we NNT mortality surveys carried out in var- selected municipalities or groups of them, ious parts of the world have provided then localities, and finally individual valuable information for decision-making dwellings (or blocks in the case of large (6, 7). Only a few such studies have been urban areas). In all, 942 Veracruz locali- reported in Latin America, but they have ties were selected, these reflecting the consistently provided higher incidence state’s distribution of urban and rural estimates (a-10) than those obtained from strata, districts (administrative units routine surveillance. One of the most re- served by a single regional office that re- cent (IO), carried out in hyperendemic port to the state government), and major areas of the Mexican State of Jalisco, doc- ethnic groups. umented a pronounced underrecording The sampling scheme empIoyed an of NNT deaths. equal probability sampling design and Cdrdems Ayala et al. NeomtaI Tetanus 117 simplified estimation procedures (13). to use a table of random numbers, divide Using this approach, assuming a fre- the city into sectors, choose one sector at quency of 2.5 NNT deaths per 1 000 live random, etc.). Interviewers (most of them births (P = 0.0025 at the 95% confidence public health nurses or community health level) with a precision of 0.001285 and an workers) received a day of training and anticipated design effect of 1.7, it was then field practice to familiarize them estimated that the households sampled with the questionnaire and interview should experience 9 860 live births dur- process. ing the study period (14). Following the sample design, the su- Death risk data were collected by ask- pervisor indicated the households to be ing mothers or guardians in the selected visited. The interviewers were instructed households “Since last Easter (April, 1988) to fill out a questionnaire for each visited has a baby been born to a woman living household. One or more full-time staff in this house?” and “Since last Easter (April, members in each local health district su- 1988) has a baby died who was born to a pervised data collection in the field. woman who lives here?” Easter being a Members of the survey group also helped fundamental event in Mexican culture, that by coding questionnaires and reviewing date was selected as a cutoff to reduce the procedures followed in the field. The recall error. It was also felt that reducing data collected were then entered into a the study period to approximately one year, database management system. as recommended in WHO guidelines, These data were later analyzed using would help improve the recalled history the point and variance estimators avail- of reproductive events (14). able in PC-SUDAAN (26). The analysis If a live birth was reported, the inform- weight assigned to each live birth was the ant was asked about the birth attendant * ratio of the estimated number of dwellings (i.e., whether he or she was a physician, from the sampling frame divided by the nurse, traditional attendant, relative, or number of dwellings in the sample; and simply the mother herself). If an infant this was used to estimate totals, under the death was reported, a verbal “autojpsy” assumption that an equal probability sam- (14, 15) was conducted in an effort to ple of households was obtained.