Two hundred eighty six infant deaths, identified by the two Death Causes Registers during the period 2004-2006 (Palermo: N = 182, Messina: N = 104), have been reviewed (Figure ). Of these, 40 (14%) were inaccurately recorded and were excluded: 26 because the infant was born to a non-resident mother and 14 because death occurred after the first birthday. The 246 death registry-confirmed cases (Palermo: N = 147, Messina: N = 99) included 143 (58.1%) males and 103 (41.2%) females, with mean age at death of 33.3 days (SD: 64.5, median: 5.5).
Cases identification algorithm and medical record retrieval.
Hospital records were not available for 68 (27.6%) of the 246 confirmed cases for the following reasons: death occurred at home without a hospital stay (N = 11); unknown place of death (N = 15); record not retrievable from hospital archive (N = 26); hospital record seized by a court (N = 8) and death occurred out of the two districts (N = 8).
The average IMR for 2004-2006 was significantly higher for the Messina district than for the Palermo district (Table ) (p = 0.0001). IMRs and NMRs were significantly higher in the Messina district both during the entire period and within each year. The difference between the two districts was statistically significant for NMRs (p < 0.0001), but not for post-neonatal MRs (p > 0.05).
Infant and neonatal mortality rates, Palermo and Messina districts, 2004-2006
The ratio of the two district-specific IMRs (RR) was 1.6 (95%CI: 1.2 - 2.1), higher for the neonatal component (RR: 1.8; 95%CI: 1.4 - 2.4) than for the post-neonatal (RR: 1.1; 95% CI: 0.7 - 1.9), indicating higher mortality in the Messina district.
The IMR was higher among male infants than among female infants both in Messina (6.8 vs. 5.0 per 1000 live births, respectively) and in Palermo (4.1 vs. 3.2, respectively). The difference between districts was evident in both genders, although larger for male infants (RR = 1.7, 95%CI: 1.2-2.3) than for female infants (RR = 1.6, 95%CI: 1.0-2.3) (data not shown).
The IMRs declined from 2004 to 2006, but the linear trend test did not achieve statistical significance for either district (Palermo: p = 0,1; Messina: p = 0,27 - data not shown). Thus, it is not possible to reject the null hypothesis that the IMR was stable during the limited time period evaluated.
Comparison of the 178 cases with available hospital records with the 68 cases whose records were not available showed no statistical difference with respect to sex (p = 0.20), year of death (p = 0.75) or district of residence (p = 0.64). Access to hospital records was possible for 77.9% of the neonatal cases, but only for the 56.9% of the post neonatal deaths (p = 0.001). This difference was independent from the district of residence (p = 0.95) (data not shown).
Among cases with available hospital records, there was a statistically significant difference between the two districts according to the distribution of cause of death categories (p = 0.02): an excess of deaths for malformations and congenital diseases in the Palermo district and an excess of deaths for preterm delivery and prematurity in the Messina district (Table ).
Characteristics of 178 infant deaths with medical record information, by district, 2004-2006
The average maternal age of infants who died in the Messina district (33.1) was significantly higher (p = 0.04) than the maternal age of infants who died in the Palermo district (31.1); there was an excess of infants born to mothers in the age categories "30-34 years old", "35-39 years old" and "> = 40 years old" in the Messina district (p = 0.02).
A NICU was present at the hospital of delivery more often for infant deaths in the Messina district (89.7%) than in the Palermo district (78.6%), but the difference did not achieve statistical significance (p = 0.06). The deceased infants were admitted to a NICU slightly more often in the Messina district (97.1%) than in the Palermo district (90.1%), but the excess was not statistically significant (Fisher's exact test, p = 0.13).
The association between selected variables and district of residence among infant deaths was evaluated (data not shown in detail). The odds of malformation being reported as the cause of death were twice as high in the Palermo district than in the Messina district (OR = 2.2; 95%CI: 1.1 - 4.3), whereas, taking "Other diseases" as the reference for "Aggregated death causes" category, the odds of preterm delivery and prematurity being reported as the cause of death were three times higher in the Messina district than in the Palermo district (OR = 3.1; 95%CI: 1.1 - 8.7).
Considering the maternal age category "<25 years old" as the reference, the odds of Messina district deceased infants increased by two or three times in the categories "35-39" and "> = 40 years old" respectively. However, the confidence intervals for all category-specific OR estimates were wide and included the null value.
In order to better evaluate the association between maternal age and infant deaths in the two districts, and understand whether the association with maternal age was a simple reflection of the older age of all mothers of live born infants among Messina, we have checked Central Institute of Statistics estimates of the numbers of infants born in each district during the three-year period of interest, stratified by maternal age. Messina district mothers tended to be older than mothers in the Palermo district (p < 0.0001). In addition, whereas IMRs increased with increasing maternal age in both districts (IMR "35-39": 3.2 Messina versus 1.9 Palermo; IMR "> = 40": 13.5 Messina versus 5.6 Palermo), the difference between the two districts increased with maternal age (Table ). Maternal age was missing for 90 deceased infants of the sample: thus, IMR estimates reported above were underestimated by about 30%. Table also displays IMR estimates corrected for missing values, assuming that the distribution of dead infants whose maternal age was unknown was the same as for deceased infants in the same district whose maternal age was available. Whereas the pattern of corrected IMRs was similar to that obtained using only available maternal age information, the corrected estimates were higher and the between-district difference in IMRs for advanced maternal age was strengthened. Overall, the data indicate that the association with maternal age observed among dead infants is due at least in part to the older age of Messina mothers as compared to Palermo mothers, but also to a particularly elevated risk of death among infants born to older mothers in Messina, as compared to Palermo. A Poisson regression model fit to the data to evaluate the IMRs as a function of "Maternal age" and "District of residence" showed a significant interaction (p = 0.04) between maternal age and district of residence, indicating that the IMRs increase with maternal age at a faster pace in Messina than in Palermo. Rate ratio estimates and 95% CIs, calculated using the age category "<25 years old" as reference (RR = 1), showed that IMRs increased about fourfold in the range of maternal age categories in Palermo, while the Messina district IMRs increased about ninefold in the same range. Thus, the model estimates that in the maternal age category "> = 40 years", the RR was 3.7 in the Palermo district (95% CI: 1.5-9.1) and 8.8 in the Messina District (95% CI: 3.7-20.7). An alternate model was fit using the corrected numerators of the IMRs described above, and yielded very similar results, confirming the larger excess mortality among infants of older mothers in the Messina district (results not shown).
Maternal age-specific IMR estimates by district of residence and infant mortality rate ratio estimates