Use of hospital-based health care services among children aged 1 through 9 years who were born very preterm - a population-based study

Children with a birth weight (BW) below 1500 g or delivered before 32 weeks of gestational age (GA) — henceforth referred to as very preterm (VPT) children — require considerable care in the neonatal period, and are at significant risk of long-term sequelae [1‐4]. The initial frailty of these children leads to poorer long-term health and social outcomes compared with children delivered at term [5‐8]. Long-term consequences have been investigated by Moster et al. [7], who described associations between levels of GA and a variety of medical and social outcomes later in childhood and young adulthood. Among these outcomes were the frequency of cerebral palsy, poor schooling outcomes, and retirement from working life due to disability.

As modern treatment increases survival at increasingly lower levels of GA, the surviving children may have higher levels of morbidity for which they require care later in life, thus studies with long-term follow-up of care provision are increasingly important to undertake [1, 9]. Even if the frequency of treatment within the tertiary care sector varied according to the characteristics of the health care system, prematurity has been shown to lead to excess hospitalizations at later stages in childhood, across a range of health care systems [10‐20].

Generally, previous studies have emphasized how variations in the use of hospital services arise predominantly in relation to GA and clinical characteristics. However, geographic, sociodemographic, and socioeconomic characteristics are known to play a role in children’s health and their utilization of health care services [8, 21]. For decision makers concerned with the long-term health outcomes associated with prematurity, monitoring the variation in hospital use can provide an indication of the persistence of initial disadvantageous clinical characteristics. Furthermore, it can serve as an indication of whether differences exist due to characteristics extraneous to clinically evaluated need [22]. The decision maker may consider expectations of equal treatment for equal need across socioeconomic and geographic characteristics relevant when considering the underlying need for care of this population.

A previous study suggested equal treatment for equal need, regardless of socioeconomic status, for preterm children remaining within the control of the health service [9]. As such, socioeconomic status should not affect the care received in the neonatal period (lifesaving initial medical treatment, respiratory support, treatment of infections, adequate nutrition or length of stay) until first discharged from the hospital. This perspective was complemented by a study by Petrou and Kupek [13], who, using older data, found more hospital admissions, longer stays, and increased hospital costs for lower social classes, defined according to occupation in the general population. These authors highlighted that the effect of social class in the general population was most pronounced in the ages 3–10 years, compared to the first 2 years of life.

The aims of the current study were to describe the relationship between age and the use of hospital-based health care services in childhood for VPT children and to investigate the distribution of health care across five sets of characteristics, namely: child health, maternal health, social background, geography, and current age.

The results of this study add to the previous literature in several respects. Nordermoen and Bratlid [2] calculated the cost of treating premature children in Norway during the initial hospital admission. We added to their results by providing knowledge about hospitalizations beyond the first life year, and consistently demonstrated that this result is more pronounced for lower gestational ages. In particular, we found that children born in the gestational weeks 22-24 and 25-26, as well as children small for their gestational age and with a diagnosis of a congenital malformation (ICD-10 section Q); have excess hospital use. In contrast to the clinical characteristics of the child, characteristics indicative of maternal health, sociodemographics, and geographical proximity to hospital services had weaker and less consistent links with hospital admissions and outpatient visits. We similarly built upon an established methodology for comparing the care outcomes for VPT children in their first year of life across countries [23, 24].

We motivated the use of imputation as a method that allowed us to retain a perspective fully compatible with a population-based study. We chose imputation as our strategy to deal with missing data on the basis that, if the variables included are predictive of missing values, then the multiple imputation procedure should aid in correcting biases and preventing the loss of efficiency that would be encountered in a complete-case analysis [29].

Other previous studies have examined utilization of hospital services beyond the initial admission [10‐20]. The present article contributes to the literature by a unique combination of register data and study design that allows us to perform a nationwide analysis of children aged 1–9 years born between January 1, 2001 and December 31, 2009. We adjusted our estimates of hospital care utilization for several measures, including maternal health, geographical access to hospital care, and socioeconomic status. Previous studies have analyzed data from specific districts, data from the 1970s and 1980s, or data covering fewer years beyond birth, or a combination of these characteristics. An extension to several previous studies is providing estimates for lower brackets of GA, and not limiting the results to a general finding for those delivered before 28 weeks. Our findings support the results of previous studies showing that clinical characteristics such as low GA and low BW contribute to hospitalizations beyond the initial admission, and that excess utilization of hospital resources on average declines with increasing age. In particular, prior to this study, hospital utilization during the period between 5 and 9 years old has been poorly documented. Similar to the study Petrou and Kupek [13], we found an increase in the effect of socioeconomic status on hospital utilization in later ages. As for immigration status, we did not uncover any effect on hospital use associated with non-native status. Disregarding the possibility of differences in cultural preferences for care at hospitals versus in primary care, this finding rejects a problem of access in case of under-utilization or problems of excessive comorbidity in the event of additional utilization compared with the native population. Acknowledging that immigration status covers heterogeneous groups we separated the growth of non-Norwegian status into that observed for mothers of European descent (increasing from 3% to 11%) and that observed for mothers of non-European descent (from 8% to 15%). This distinction emphasizes that compared to native Norwegians; both origins became more prevalent over the period considered, but that the increase was proportionally larger for mothers of European rather than for mothers of non-European descent. This situation accords with the overall development for Norway where the main increases in immigration in the past decade have come from countries in Eastern Europe.

We have documented variation in the utilization of health care resources according to place of living at birth for a nationwide population of VPT children. A recent report outlined substantial geographical variation in the use of hospital care for the general population of Norwegian children aged 0-16 years [30]. The authors of this report noted that the variation in hospital use could not be accounted for by differences in health status or by the characteristics of the health care providers. Similarly, a pattern by geographical proximity was not an apparent explanation for the observed differences. To the extent that we have adjusted the use of hospital care across hospital regions for potential systematic differences by health status; we note similar differences in the current study. Thus, the reasons for variation across regions remain unaccounted for. A question is whether this result is due to differences in local attitudes toward viability at birth. In particular, restricting care to babies born before 25 weeks GA may affect resource use later in childhood. An indication of such local attitudes may present itself by differences in the rates of live births. The Norwegian Research Council held a consensus conference in 1998 on the issue of gestational age limits for treatment of extremely premature infants [31]. This conference concluded that offering treatment to infants at 23 weeks’ gestational age was considered experimental medicine; in weeks 24 to 25 treatment could be offered if in agreement with the parent’s wishes and after week 25 treatment should be offered unless contraindicated for medical reasons. Hence, both the hospital regions as administrative entities and the therein residing parents are possible sources of local attitudes to viability. To see if this is reflected in the data we present a table on live and total births by the age of gestation and hospital region. In keeping with the categorization of gestational age in the current study, the table has been limited to infants below and above 25 weeks of gestation. Since there are four regions, we test for differences in expected versus observed live births across these using a chi-square test [Additional file 1: Table S5]. We find evidence of a difference for babies born <25 weeks gestation; in particular, for the Central health authority, whose observed live births are beyond those expected for the country as a whole. This result may indicate the presence of local attitudes to viability both from the side of parents and hospital regions, but from the size of the affected population we are unlikely to observe the effects of such sorting for hospital use of children aged 1-9. In all, we identify 530 live births of a total 1124 births with GA <25 weeks. However, only 198 are present in our analysis sample, mainly due to the exclusion criteria that limits the sample to children surviving to age one. For children age 1-9 the <25 GA group accounts only for 3.7% of our sample (cf. Table 2). Thus, the reasons for any difference between the hospital regions can be expected to lie elsewhere. It is a limitation that we are unable to ascertain why this population varies according to hospital regions in the use of hospital care. Further studies are needed to investigate the reasons for the geographical variation in use of hospital care for children born VPT and the general population of children and young adults alike.

The study had some limitations. First, owing to the study design we are unable to examine differences in hospital admissions and outpatient visits between those born early and late in the 2000’s as potential cohort effects. Generally, mortality by age one among the very preterm remained stable over the years considered as assessed by our data however other differences in the health of these children by birth year may not be captured satisfactorily by mortality as an indicator. Second, although it is generally a good practice to model continuous variables as such, we chose to model continuous age in two categories, 1-4 years and 5-9 years. Our choice was motivated by providing results that are easy to interpret and conforming to what we perceive as common categorizations in the literature. A more flexible modelling of continuous variables with for example splines might have been preferred methodologically, but would also complicate a presentation of results and discussion of possible developments with age by other characteristics. In future work, we aim at introducing a more flexible modelling of continuous variables to our work. Third, only admissions and consultations at public hospitals were included. For hospital admissions this is not a problem, since all admissions for this group of patients are at public hospitals. However, a considerable proportion of outpatient consultations are provided by privately practicing physician specialists who contract with a regional health authority. Iversen and Kopperud [32] showed, based on general population survey data, that socioeconomic status is positively correlated with utilization of private specialists once self-assessed health status has been adjusted for. If this result is generalizable and applicable to the population of VPT children, we may have underestimated the effect of socioeconomic status on health care utilization in this study. The effect of socioeconomic status on hospital care utilization is worth a discussion. As a measure thereof, we used maternal grandparents’ education in preference to the education level of the mother. In our data, the education of the grandparents is explanatory of the mothers’ attained education when adjusted for age and parity of the mother in a separate regression model. (The prevalence of mothers with higher education increases 33 percentage points if the maternal grandparents are also highly educated (CI 30-36).) We made two observations in models equal to those presented in Table 3 when we included the mother’s attained education in addition to grandparents’ education or mother’s education only. First, the coefficients of grandparents’ education were slightly stronger when included alongside maternal education, exhibiting the same pattern and statistical significance for the older age group. Second, the coefficients of maternal education were small and insignificant for either age group. Thus, while the two indicators of socioeconomic status are related, it is maternal grandparents’ education that contributes to explaining the use of hospital-based care. In general, an association between parents’ socioeconomic status and children’s hospital utilization can be established through at least two channels. First, parents with high socioeconomic status may communicate more easily with health professionals and, hence, be able to obtain better treatment for their children. This effect has previously been found by Finnvold [33]. Second, children of parents with high socioeconomic status may have better health status, as found by Petrou and Kupek [13]. This may not have been accounted for fully in the register data we used. Better health may imply less need for health care. Our empiric result of no association between parents’ socioeconomic status and children’s hospital utilization in the younger and needier age group shows the total effect of the two channels, and is a weak basis for making normative judgments.

The relevance of this nationwide study is both related to exposing the variables that impact on hospital utilization of VPT children throughout ages 1–9 years and to predicting future hospital utilization for this patient group. We demonstrated substantial differences in hospital care utilization across age groups and clinical factors, adjusted for a host of factors pertaining to geographic and social background, and maternal health. The added information from these latter factors, although limited, was somewhat more evident within the 5–9 year age group (for social background), and mixed for geography and maternal health. The use of hospital care for the population of VPT children remains higher than their non-VPT peers in the general population throughout the ages 1–9 years.

Figures 4 and 5 show how care is distributed as a fraction of the population. The result was that the great majority of the children was healthy beyond the first year of life and did not require any particular follow-up from the hospital sector. Between 1 and 4 years of age, 22% of VPT children had at least one hospital admission. This reduced to 13% for 5–9 year-old children. As expected, inpatient admissions were more concentrated than outpatient consultations. This means that outpatient consultations were more evenly distributed among the children than inpatient admissions. Embedded into the graphs are figures of the distribution of use of the type of care for those with at least one hospital contact. These embedded figures demonstrate that, while fewer 5–9 year-old VPT children are in contact with the tertiary health care services, those who do remain have a similar distribution of care as in the 1–4 year age group.