Estimating the cost of congenital Zika syndrome to families and healthcare providers in Rio de Janeiro and Pernambuco, Brazil: results of a case-control study [version 1; peer review: 2 approved with reservations]

Background: Children with congenital Zika syndrome (CZS) have a wide range of additional healthcare needs. This study aimed to estimate the direct costs of CZS from the health provider and family perspectives, and the indirect costs for families, in two Brazilian states: Rio de Janeiro and Pernambuco. Methods: A case-control study was undertaken between May 2017January 2018 recruiting 174 cases with severe CZS, 41 with mild/moderate CZS and 269 children with no CZS, across the two sites, from existing studies. The primary caregiver was interviewed using a structured questionnaire to collect information on healthcare use and costs incurred during the previous 12 months. In Rio de Janeiro, health care utilization data was also extracted from electronic medical records. We estimated direct and indirect costs incurred as a result of CZS from the perspective of the health system and families. Results: Children with CZS accessed more healthcare facilities and reported longer travel and waiting times than children unaffected by CZS. Total costs from the health provider perspective of outpatient visits, were highest for children with severe CZS (U$1,411) followed by children with mild/moderate CZS (U$264) and children without CZS Open Peer Review


Introduction
In 2016, the world health organization (WHO declared the Zika virus (ZIKV) epidemic a "public health emergency of international concern", because of its association with a wide spectrum of congenital anomalies, now collectively called "congenital Zika syndrome" (CZS) 1 . Brazil was the most severely affected among the countries and territories reporting outbreaks; by 2018 there were 2,952 cases of "confirmed congenital syndrome associated with Zika virus infection" in Brazil, making up 79% of cases for the PAHO region 2 . The true number is likely to be far higher. ZIKV spread throughout all Brazilian regions, but the epicenter of the epidemic was in the Northeast region 3,4 .
Microcephaly is the most severe CZS complication in children, but congenital infection with ZIKV is also associated with a range of other conditions, including hypertonicity, seizures, ophthalmic abnormalities, arthrogryposis, cardiac conditions and early development delay [5][6][7] . Consequently, children with CZS are likely to experience a wide range of additional health and social needs during their lifetime, including for specialized health care and education 8 . Moreover, families of children with CZS can experience substantial financial impacts, as they try to meet the needs of their child, for instance paying for travel to receive healthcare or purchasing assistive technologies. Additionally, mothers often assume the role of the main caregiver and may not return to employment 9 .
The outbreak of ZIKV in Brazil occurred in the middle of an economic crisis that began in 2014 and resulted in a set of fiscal austerity measures implemented from 2016, including freezing the social and health care spending 10 . The national health system (Sistema Único de Saúde -SUS) provides universal health care with its services financed and delivered at the federal, state and municipal level. While total public health expenditure per capita by municipalities grew by 226% from 2003 to 2014, it has decreased by 6.3% since 2015 11 . Provision of efficient and effective healthcare services tailored to children affected by CZS is therefore challenging for the SUS, due to the funding constraints and the high and diverse needs of the children for public resources, including for a broad range of medical specialties and technologies.
To date, the economic impact of CZS for families and the health provider is unknown. One model created to inform a cost-effectiveness tool, by AlfaroMurillo et al., predicted that each case of microcephaly incurs direct medical costs of $91,102 and $28,818 over the lifetime for Latin America and the Caribbean, respectively 12 . However, these results were estimated by extrapolating the costs of a case of intellectual disability from the United States, and actual data has not been generated for the Zika epidemic. Other analyses have focused on the economic burden of microcephaly in the USA, but the findings may not be relevant to Brazil or other Latin American settings 13 . The lack of data on economic impacts of CZS is an important gap, as understanding the social and economic consequences of CZS in Brazil could help decision-makers in allocating optimal resources to support these children's healthcare and social needs.
The aim of this study was to estimate the direct cost of caring for a child with CZS from the health provider perspective and the direct and indirect cost for families over a 12 month period, in two Brazilian settings: Rio de Janeiro city, capital of the state of Rio de Janeiro and in Recife city, capital of the state of Pernambuco.

Study setting
A case-control study was carried out in two contrasting sites in Brazil, selected from where research was ongoing and the teams had good access to families of children with CZS 14 . The first was Recife city and Jaboatão dos Guararapes, in the State of Pernambuco in Northeast Brazil. The Northeast region has a high number of suspected and confirmed cases of CZS and is considered the epicenter of the epidemic. For contrast, the second site was Rio de Janeiro city, in the state of Rio de Janeiro, where symptomatic ZIKV was less prevalent and reports of CZS far lower. Collectively, the States of Rio de Janeiro and Pernambuco accounted for 25% of confirmed CZS cases in the country 4 . Furthermore, Rio de Janeiro and Pernambuco have among the largest public health care networks in Brazil, with most of hospitals and outpatient facilities linked to the SUS 15,16 . The protocol of the study has been published in full previously 14 .
Children in the study were born from late 2015 through 2016. Economic data (health provider costs, and direct and indirect costs incurred by the families) was collected between May 2017 and January 2018 through use of a questionnaire completed by the child's caregiver and electronic medical records (Rio de Janeiro only).

Cases and controls
In Pernambuco, the majority of cases and controls were identified from an existing case-control study initiated in January 2016 17 . In Rio de Janeiro, the main source of the cases and controls was from an ongoing cohort study.
Case selection. In Pernambuco, cases were children born with microcephaly, head circumferences < 2 standard deviations (SD) than the mean in eight public maternity hospitals. Additional cases with microcephaly were identified from an ongoing cohort of pregnant women who presented with a rash (a common symptom of ZIKV infection), and from outpatient clinics of children with CZS (mostly from Oswaldo Cruz hospital). They were classified as severe or moderate CZS, based on their head circumference ("severe" head circumference <3 SDs below the mean for age and sex).
In Rio de Janeiro, the source of the cases was the Vertical Exposure to Zika Virus and Its Consequences for Child Neurodevelopment: Cohort Study in Fiocruz/IFF (Clinical-Trials.gov Identifier: NCT03255369). Cases were children born to mothers known to be ZIKV positive, who either 1) had microcephaly or significant developmental delay (i.e. had a composite score <70 on the Bayley Scale of Infant Development scale 18 between 6 and 36 months) and presented with other clinical conditions with eye or hearing abnormalities or other brain malformations ("Severe CZS"), or 2) had less severe developmental delay indicated by a composite Bayley score of 70-84 ("Mild/moderate CZS").

Control selection.
In Pernambuco, controls were children born in the same hospitals, but without microcephaly and without neurological or other health problems (determined from transfontanelle ultrasonography, and through physical examination by the study neonatologist), with both examinations performed soon after birth 17 . Controls were matched to cases on the basis of expected date of delivery and place of mother's residence (by health region). During the follow-up interview in 2017/2018, parents were asked whether there were any developmental delays (using the Denver Developmental Screening Test) 19 , and if the response was positive, they were excluded from the study and referred for further investigation.
In Rio de Janeiro, control subjects were born to mothers without a history of symptoms and without developmental delay, as shown by: 1) a composite Score ≥85 on the Bayley Scale of Infant Development scale 18 , conducted between 6 and 36 months following the recommended guidelines and/or 2) assessment by two paediatricians based on the child's medical records 14 . In Rio de Janeiro, the sample of controls included nine pairs of twins and, for each pair, one child was randomly selected for inclusion as a control, in order to avoid double-counting of families.
According to case and controls definitions, children in both settings were categorized into three groups: (1) children with severe CZS (microcephaly or with serious developmental delay); (2) children with mild/moderate CZS; and (3) children not affected by CZS.

Sample size justification
The sample size for the overall study was powered to detect a difference in depression prevalence between the control and case mothers, as this was one of the study's primary outcomes 14 . We aimed to recruit 100 cases and 100 controls per setting, which would provide the power to detect an odds ratio (OR) of 2.6 in each site for the association between depression and CZS, assuming 95% confidence, 80% power and a prevalence of depression of 15% in unaffected mothers. Across the two samples (i.e. 200 cases and 200 controls), the sample size would be adequate to detect an OR of 2.1 for the same association. We considered that the sample size would be sufficient for the current analyses, given the large expected economic impacts of CZS.

Data collection
Two sources of data were used to estimate the direct and indirect costs of CZS.
First, the primary caregiver (usually the mother) was interviewed using a semi-structured questionnaire, between May 2017 and January 2018 (full questionnaire available as extended data 20 . Data were collected on socio-economic status, clinical data, direct costs (e.g. travel, food, parking, hospital/ exam fees) and indirect costs from families and main caregiver (e.g. productivity loss/opportunity cost), health care resource utilization (frequency of outpatient visits to different types of health professionals, and number of hospitalizations), and coping strategies (selling of assets, borrowing money). The effort of the caregiver in seeking care was also estimated by asking about the number of healthcare facilities where the child received care, transportation time and waiting time at the health care facility for a typical visit. The reference period was the last 12 months. The cost component of the questionnaire used a tool developed by the UK working party on patient costs as a template and adjusted it to the context 21 . Additionally, the questionnaire was used to collect data from the caregiver on the family monthly income and other sociodemographic characteristics, as well as parental indicators (e.g. age, marital status, schooling, depression, anxiety and stress, social support). For analysis, monthly income was converted into yearly income.
The second data source was the CZS cohort´s electronic medical records database that recorded resource utilization (hospitalizations, tests and exams) for each child in the Rio de Janeiro sample, between June 2016 and April 2018. Hospital data from electronic medical records that included hospitalizations, tests and exams were not available for Pernambuco.

Calculation of health provider's costs
Costs from the national health system perspective were calculated using an ingredient approach for each child and each year, whereby each resource used was identified, quantified and valued. The number of consumed resources (outpatient visits, hospitalizations, and tests) were multiplied by the costs on the national cost reference table and adjusted by a factor of x3.51, as suggested in the literature 22,23 . This factor was estimated by an analysis of the cost of some procedures, as the national cost reference table did not reflect the real cost of these procedures as it had not been updated for approximately eight years.
The number of consumed resources for outpatient visits, hospitalization and tests and exams were estimated as follows: -Medical appointments: The number of outpatient visits were collected from the questionnaire for both settings and included a broad range of specialties (e.g. genetics, neurology, nutrition, occupational therapy, ophthalmology, orthopaedics, paediatrics, physiotherapy, and audiology). The costs of these were estimated from the national cost reference table, with adjustment.
-Hospitalizations: The number of hospitalizations were collected from the questionnaire. For Rio de Janeiro, the electronic hospital database contained information on all procedures undertaken for these hospitalization events, which included surgical (e.g. gastrostomy tubes) and non-surgical procedures (e.g. pneumonia treatment). These interventions were costed using the national cost reference table, with adjustment. We assumed that the mean medical direct cost of hospitalization by group of cases in Rio de Janeiro was a proxy of the cost in Pernambuco.
-Tests and exams: Tests and exams received were available for the children in the Rio de Janeiro arm of the study from the electronic hospital database, and these included reverse transcription polymerase chain reaction (RT-PCR), eye examination, hearing examination, fat and free mass composition, composite Bayley test 18 , laboratory tests, and brain images (transfontanelle cerebral ultrasonography, computed tomography, and magnetic resonance imaging). The costs of these tests were estimated from the national cost reference table, with adjustment, except for the costs of RT-PCR, fat and free mass composition and composite Bayley test which were not included in the national cost reference table. Therefore, micro-costing was applied to estimate their costs using inputs from the from Brazilian government price databases and local suppliers. We did not include test and exam costs of children in the Pernambuco arm of the study.
All costs were converted into U$ 2017 at the exchange rate of 3.19 Brazilian reais/U$ (www.http://data.imf.org). No discounting or inflation rate adjustments were applied. Costs were annualized when necessary.

Calculation of family costs
Direct, and indirect costs incurred by the family were collected through the questionnaire.
-Direct costs included all out-of-pocket expenditure (food at healthcare facility, drugs, special milk formulas, glasses, and other health care expenses, as well as transportation, costs associated with family displacement from their home and the renovation of the patient's home). Costs were defined as catastrophic expenditures if total direct costs were equal or above 40% of yearly household income (estimated using the Critério Brasil based on assets and infrastructure of the household) 24 .
-Indirect costs included the productivity loss of the caregiver and were estimated as the days of work lost to take care of the child, the time invested in transportation and the waiting time seeking for health services. The value of one workday was estimated using the median of household income provided by Critério Brasil divided by the number of persons living in the house and monthly working days. Coping costs (selling of assets and borrowed money) were measured through the questionnaire in order to understand how families handle these expenses.
Costs were calculated for a 12-month period. For children aged less than 12 months at interview, an extrapolation was made for the full 12 months based on monthly consumption data available.

Data analysis
Median and interquartile range (IQR), for continuous data and frequency distributions for categorical variables were used to describe the study sample, effort of main caregiver, and direct and indirect costs, all stratified by age, location and group (severe CZS, mild/moderate CZS no CZS).
Cost data was statistically analysed using the Kolmogorov-Smirnov test to assess the normality assumption. Differences between groups were verified through a pairwise comparison using the student's t-test and the Wilcoxon-Mann-Whitney. We also used chi-square test to compare categorical variables of effort of main caregiver between the groups. A 5% level was chosen as the level of significance. Data management and statistics analysis were performed using R (R Core Team. All interviewees were adults and provided written informed consent, as outlined previously 14 .

Characteristics of study subjects
A total of 484 main caregivers were interviewed, 56% from Rio de Janeiro and 44% from Recife (Table 1). Of the 484 children, 36% (n=174) had severe CZS, 9% (n=41) mild/ moderate CZS and 56% (n=269) were children not affected by CZS. The sample was made up of approximately equal numbers of boys and girls. Most of the children (70%) were more than one year old at the time of the interview. In Rio de Janeiro the children with severe CZS were on average younger than the those with no CZS, while in Pernambuco the opposite was true. Approximately half of the children lived outside of the cities of Rio de Janeiro and Recife. Mothers were the main caregivers for more than 93% of children.
Overall and in Rio de Janeiro caregivers of children with severe CZS were generally younger than control caregivers, while in Pernambuco the groups were well matched on age. A high proportion of all caregivers lived with their partner (75%) and had nine or more years of education (82%). Overall, and in Rio de Janeiro, households of children with severe CZS were more likely to belong to the lowest socioeconomic class (39%) compared to controls (30%), but this difference was not observed in Pernambuco. The majority of households of children with severe CZS were receiving social benefits (62%), while this was less common in households of children with mild/moderate (34%) or no CZS (29%). Receipt of social benefits was more common in Pernambuco than Rio de Janeiro, across all three groups. Households were relatively similar in terms of household size and number of children, regardless of location and CZS status. Effort in seeking healthcare by the main caregiver In the previous year approximately half of children with severe CZS (56%) and with mild/moderate CZS (44%) had to access at least three health care facilities, while utilization of three facilities was much rarer among children with no CZS (7%) ( Table 2). Time taken to reach the healthcare facility for families of children with severe CZS was frequently more than three hours (46%), while this proportion was lower among children with mild/moderate CZS (20%) or no CZS (28%). Waiting time at the health care facility was often more than three hours, across the three CZS groups. These patterns were consistent in both Pernambuco and Rio de Janeiro, except effort scores were particularly low among the "no CZS" group in Pernambuco.

Health provider costs
The median health provider cost of outpatient visits for children with severe CZS (U$1,411, IQR 740-2,060) were significantly higher than children with mild/moderate CZS (U$264, 174-1,144, p<0.01) and those not affected by CZS (U$107, 74-209, p<0.01) (  Table 4 shows the direct and indirect costs incurred by families for each group of children. Total direct costs for food, transport or other items, were consistently higher for children with severe CZS (total U$1,129) compared to those with mild/moderate CZS (total U$231), which in turn were higher than for children with no CZS (total U$38). These patterns for direct costs were consistent in Rio de Janeiro and Pernambuco, although costs were generally higher in Rio de Janeiro, in particular transport costs. Median indirect costs were generally low. Productivity costs were higher for families with children with CZS compared to children without CZS. Stratification by location revealed more substantive indirect costs associated with Severe CZS in Rio de Janeiro, but that in the other groups and in Pernambuco these costs remained low.

Discussion
The main purpose of our study was to estimate the direct and indirect costs of caring for children with CZS, from the family and health provider perspective, based on a study carried out in Rio de Janeiro and Pernambuco. This analysis was based on three groups of children, selected to represent the broad     IQR, Interquartile range. NA, not available. This information was not collected for children without CZS spectrum of clinical presentations of this condition: children with severe, mild/moderate and no CZS. We also examined in depth the effort of the main caregiver in their interaction with the health care system, as children with CZS had a high consumption of resources, likely to continue throughout their lifetime. To the best of our knowledge, this was the first study to quantify the costs incurred from caring for a child with CZS from the health provider and family perspectives using primary data.
This study showed that children with severe or mild/moderate CZS accessed more healthcare facilities and reported longer travel and waiting times than children unaffected by Zika.
Costs of outpatient visits were highest for children with severe CZS followed by children with mild/moderate CZS and children without CZS, although there was some variation between the two settings. Direct and indirect costs incurred by families were higher for families of children affected by Zika, in particular as a result of "other" costs, such as drugs and special milk formula.
Most of the main caregivers of children with CZS were mothers who experienced high time costs in seeking and accessing care for their children. One contributing factor to this high effort is that almost half of cases lived outside the metropolitan area of Rio de Janeiro and Recife and so healthcare services were situated far from households in both cities. The transportation costs incurred by caregivers were much higher in Rio de Janeiro than in Pernambuco, potentially because in the latter setting caregivers had access to public services provided by government to allow them to reach the healthcare facilities. Thus, given such high care needs in a context of fragmented health services provision, the coordination of the health system would be very helpful to caregivers, saving time and money 25 .
Our analysis also shows that many families of children with CZS in both settings faced catastrophic expenditures, using a threshold based on a multi-country study, that included Brazil 24 . Catastrophic health expenditures had a high potential to lead to impoverishment, considering that about 40% of families of children with severe CZS reported an average monthly income of U$ 240.75. This situation was exacerbated by the Brazilian economic crisis that has been experienced since 2014, through increase in unemployment and austerity measures leading to a reduction in social expenditures and cuts in the health budget. According to Brazilian public policies, these children should be guaranteed access to health care services and transport systems. However, the high costs incurred by families of children with CZS in transport and drugs show that this has not happened. The impoverishment of households in Brazil due to costs of drugs has been demonstrated previously for other conditions 26 . It is also apparent that families of children with CZS should be given social benefits in order to reduce the financial impact experienced, yet at the time of the study only 62% of caregivers of children with severe CZS and 34% of those with mild/moderate CZS received social benefits. Enhancing and coordinating social and childcare policies is essential to support families in coping with the situation.
Another concern is that there are still gaps in services for children with CZS. Between 2015 and 2018, about 79% of children with CZS in Pernambuco had access to specialized health care services but only 35% in Rio de Janeiro 4 , and nutritional services are also reported to be inadequate for these children 27 . Gaps in access to required healthcare services pose specific issues that need to be addressed by policymakers, as they highlight the need for a well-structured health care network that caters to all levels of assistance and avoids caregivers of children with CZS having to access many different facilities to seek care. It is likely that provision of these services will push the costs of caring for a child with CZS higher still but may also improve health status and thereby protect from other health care costs.
Meeting these costs will be challenging, since the current economic crisis in Brazil has meant that federal health spending will be limited over the next twenty years. The annual SUS budget was approximately USD $62 billion (USD$, 2017) before the funding process changed in 2018. Early estimates suggest that the SUS budget will suffer a cumulative loss of around U$190 billion between 2017 and 2036 28 . Thus, the states and municipalities may need to increase their share of responsibility for funding SUS, although this seems difficult as the per capita expenditure has been declining since 2015 11 . The economic crisis and austerity measures have been shown to impact negatively on social policies and contribute to the deterioration of health care systems offering universal coverage 29 . This issue will be affected still further by the ongoing COVID-19 pandemic. The Brazilian health care system is facing scarcity of resources, inefficient resource allocation and uncoordinated care that could worsen the health outcomes of most vulnerable populations, such as children with CZS, even further.
There are few studies that provide cost analyses of CZS in countries where ZIKV have been recorded and that would allow comparison with our findings. A recent economic analysis in the US used the productivity costs associated with autism as a proxy to estimate the costs of microcephaly. However, results were likely to be substantially underestimated because the costs associated with autism are lower than those of microcephaly 30 . In another study, the lifetime medical costs of microcephaly to selected countries of Latin America and Caribbean amount to U$180,004 per case (in 2015 U$) and considered the costs associated with severe intellectual disability in the US in 2003 31 . Both studies had to use assumptions to calculated costs of microcephaly because there is a dearth of previous estimates of the economic impact of microcephaly in many countries, and they lacked the primary data that we collected in the current study. A third study undertaken in Texas, USA also estimated high economic costs from inpatient hospitalization for babies with microcephaly, which may provide insights into a potential impact of Zika. However, it is hard to infer costs from USA to Brazil or other Latin American settings 13 .
There are several limitations of this study. The study was conducted in only two settings and therefore does not fully represent the entire country. Nevertheless, these states included 25% of confirmed and suspected cases of CZS in Brazil 4 . In Pernambuco and Rio de Janeiro, attempts were made to include all eligible children with severe CZS, but inclusion of children with mild/moderate CZS was less comprehensive and so they may not have been representative of cases in this category overall. Additionally, the sample selection approach varied in the two sites, which may have influenced comparability across Pernambuco and Rio de Janeiro. For instance, in Pernambuco the cases and controls were matched on location while in Rio de Janeiro they were not, which explains in part the greater socio-economic differences between these groups in Rio de Janeiro than Pernambuco. Consequently, cases in Rio de Janeiro may have been more at risk of catastrophic expenditure than controls. Notwithstanding these differences, the expenditure by cases in Rio de Janeiro (whether severe CZS or mild/moderate CZS) were greater than in Pernambuco, and the difference in catastrophic expenditure between severe and no CZS groups similar in the two settings. Another limitation is that we did not consider costs associated with tests and exams in Pernambuco, due to the lack of hospital electronic medical record data, and this may bias our cost results. Furthermore, children were classified as having "no CZS" on the basis of tests and reports from when they were under two years of age. However, developmental delays may have become apparent as the child became older, and there is increasing evidence that congenital infection with Zika may cause neurodevelopmental implications in children without CZS 32 . Consequently, some of the children in the "no CZS" group may have additional needs, and this may underestimate the true economic impact of CZS.
It is important to note, that while this study focused on economic costs, caring for a child with disabilities can have many positive outcomes. Furthermore, children with disabilities have fundamental human rights, including access to healthcare. Our purpose in this paper is to encourage appropriate budgeting and planning of services, and not to put into question the need for healthcare investment in children with disabilities.

Conclusions
Children affected by CZS incurred greater costs, both from the perspective of the health provider and the family. This study supports the need for additional public health resources to meet the needs of those affected. Moreover, the magnitude of costs attributable to CZS for the families affected, and the high levels of catastrophic expenditure, shows a need to develop health related financial risk protection initiatives for households of children with CZS, and other comparable health situations.

Data availability
Underlying data Data associated with this study will not be made freely available, owing to the small number of children with CZS, making data potentially identifiable, and the sensitive nature of the subjects discussed in the interviews. However, we are committed to collaborating with other researchers in the analysis of our data (full questionnaire available online) 14 . Applications for access to the raw data for this study should be made by contacting Professor Hannah Kuper (hannah.kuper@lshtm. ac.uk), Dr Tereza Maciel Lyra (terezalyra@cpqam.fiocruz.br) or Dr Maria Elisabeth Lopez Moreria (bebeth@iff.fiocruz. br) and outlining the purpose of the proposed analyses and the variables requested. These applications will be reviewed by the three researchers, and if accepted, the requested variables will be shared. references.
If possible, expand more in the discussion regarding what were the cost factors that contributed most to the economic burden (i.e. transport or food etc). Consider expanding on the "other" cost category for table 4 in the methods.
I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.