Caregivers of children with health problems (CHPs) experience a broad range of challenges that extend well beyond typical parenting responsibilities.1,2 These challenges include physical and psychological health problems that occur more frequently than among caregivers of children without health problems.2–7 Importantly, this issue does not appear to be limited to specific types of child health problems, nor is it limited to the most severe cases. Large-scale survey work has shown that even when child health problems are defined broadly and include nearly one-quarter of the population, caregiver health problems are significantly higher.6,8 The prevalence and range of these health problems make this an important public health issue that requires attention.
Efforts to describe and better understand caregiving-associated health problems across populations have until recently been limited to cross-sectional data, which are limited in the questions they can answer. Such studies cannot address time-dependent questions important to understanding causal relationships and informing preventative interventions targeted at families caring for children with health problems. Do caregiving-associated health problems originate with the child with health problems, or do they exist before childbirth? How long do these health effects last, given some work suggesting that they can be short-lived9 or last for years?10 Does caregiver health get worse over time, as suggested by some work on caregivers of patients with Alzheimer disease,11 and some population-based survey work,10 remain stable,12 or actually improve, given positive and protective aspects of caregiving in some situations?13 Answers to these questions could inform the development of supports and enable more targeted public health interventions by identifying when such interventions might be most effective. To date, no study has examined changes in health of a population-based cohort of mothers before, during, and after the birth of CHPs using linked administrative health data.
This study employs a unique, population based, linked administrative health dataset to examine the health of mothers (ie, the primary caregiver in over 90% of Canadian families with a child with a health problem)2 over an 11-year period before and after the birth of a child with health problems. Using administrative health data on all mother-child dyads for children born in the year 2000 in British Columbia, Canada and sorting mother-child dyads according to previously developed, noncategorical definitions of child health,14,15 we are able to create health trajectories for mothers of CHPs, and compare them with mothers of children without health problems (non-CHPs). This approach allows us to establish whether caregiving-associated health problems existed before the birth of the child, whether they are particularly strong in the perinatal period, and how they change over the early childhood period.
Study Population and Data Sources
Canada’s westernmost province (pop. ~4.6 million in 2014), British Columbia’s universal health care is administered by the provincial ministry of health through programs like the Medical Services Plan (MSP), which covers physician services, and PharmaCare, which provides prescription drug insurance. We analyzed data from children born in the year 2000 and who, with their mothers, were enrolled in the British Columbia MSP from 2000 to 2007. As the study assesses the health of mothers over time, only mothers who were continuously registered with the MSP every year between 1996 and 2007 were included in the study, which corresponds to 68% of all mothers of children born in 2000. Children born in the year 2000 were selected in order to have a sufficient number of both prebirth and postbirth years of data. At the time the data were obtained (September, 2012), pharmaceutical, medical services and hospitalization data were available from January 1996 up to March 2008. Selecting a cohort born in the year 2000 allowed us to categorize children by health problems at age 7, as many developmental problems are diagnosed at the age of school entry.16 This time frame also allowed for an assessment of prebirth trajectories starting 4 years before the child’s birth.
Data were obtained from Population Data BC (PopData), a province-wide data repository that houses one of the world’s largest collections of linked health care, health services, and population health data. This study uses data from 5 distinct administrative data holdings:
- The MSP Payment Information File:17 medical services provided by fee-for-service practitioners to people covered by BC’s universal health insurance program.
- The Discharge Abstract Database (DAD):18 hospital discharges, transfers, and in-hospital deaths of patients from all of BC’s acute-care hospitals.
- PharmaNet:19 prescription data for drugs and medical supplies (eg, insulin pumps, orthotics) dispensed by pharmacies.
- The Consolidation File (MSP Registration and Premium Billing):20 population demographic data prepared for research use by PopData.
- Aggregated Diagnosis Groups (ADGs) data: generated by PopData using the Johns Hopkins Adjusted Clinical Groups (ACG) Case-Mix System software version 10.0,21 described in more detail below.
All PopData data are linked using encrypted, randomly assigned identifiers that match personal health numbers across databases, and thereby provide individual level data that are de-identified to maintain anonymity and confidentiality.
Data access was approved by Data Stewards at the BC Ministry of Health, and ethics approval was granted by Ottawa Health Science Network Research Ethics Board, where the research was carried out. The encrypted data files were made accessible to the research team on a secure research environment through PopData.
Child health was assessed in 2006 based on 2 previously developed, noncategorical approaches to identifying CHPs.14,15 One indicates whether the child had high service use based on 2 items (prescription medication and service use) on the Children with Special Health Care Needs (CSHCN) Screener,22 the other indicates a diagnosis of a Major and/or Chronic health condition. Diagnosis of a Major and/or Chronic condition was defined by the Johns Hopkins ACG Case-Mix System.14 High service use was defined as a high (ie, above the 95th percentile) number of physician visits, or medication use for 9 of 12 months.14 We should note that the other 3 items of the CSHCN Screener (ie, functional limitations and the use of special therapies and counseling services) were omitted in this study because they are not well-recorded in provincial administrative data. For these analyses, we focused on dyads where children met both high service use and diagnosis (CHPs) or neither criteria (non-CHPs).
Mother/Child Dyad Selection
Children were linked to mothers based on the MSP contract number, which is the same for all family members. Only children linked to one mother were included (88% of cases; children could be linked to more than one mother in cases of foster, marital separation, new family formation, etc.). For mothers linked to more than one child born in 2000, 1 child was selected, with priority given to any child meeting the CHPs criteria. A variable was also created to indicate whether or not the mother had an older child meeting the CHPs criteria.
Maternal health was assessed with 3 distinct outcomes, examined between 1996 and 2007, from 4 years before the child’s birth until 7 years afterwards, with the specific timing of each year based on the child’s birth month (eg, for a child born in July 2000, the year before birth would be considered July 1999 through June 2000). All 3 outcomes have been used previously as indicators of health; number of physician visits and prescriptions per year have been used in previous research in both pediatric and adult populations,14,15,23 and administrative data has been shown to be valid in terms of identifying adults with chronic conditions.21,24
Number of Maternal Physician Visits Per Year
Physician visits were based on the number of unique records by date and specialty, excluding healthy pregnancy-related or birth-related and lab/x-ray visits, as the goal was to assess health problems rather than normal pregnancy-related services. Specific ICD-9 codes for the excluded visit types are listed in Appendix A, and their description is presented in Appendix B (Supplemental Digital Content 1, http://links.lww.com/MLR/B752).
Number of Maternal Prescription Drugs (Level-3 Anatomic Therapeutic Classification Codes)
Prescription medications in the PharmaNet data are all assigned an Anatomic Therapeutic Classification (ATC) code, which divides active substances into groups according to the organ or system on which they act, and their therapeutic, pharmacological, and chemical properties. Third-level ATC codes represent major therapeutic or pharmacological subgroups (eg, C07A are beta-blocking agents; R03A are adrenergics, inhalants). The number of different medication types is a count of unique level-3 ATC codes.
Selected Physical Maternal Health Conditions
We identified whether mothers had 1 or more of the following diagnoses in a given year based on ICD-9 codes (listed in Appendix Table B, Supplemental Digital Content 1, http://links.lww.com/MLR/B752): allergy, arthritis, asthma, back problem, bronchitis, cancer, diabetes, hearing problem, heart disease, herpes, hypertension, injury, migraine, sinusitis, ulcer, glaucoma, and cataracts. These selected health conditions were selected to align with previous work based on Canadian population survey data.2,6,8,25,26 This list of outcomes was originally intended to include a broad (but not exhaustive or specifically caregiving-sensitive) selection of maternal health conditions. For this study, we use similar outcomes to align with previous research and discuss (post hoc) findings based on self-reported versus administrative data.
Several variables were used to assess how maternal health changed over time. The temporal predictor time (coded as integers 0–10) represents an increase of 1 year, from 4 years before the child’s birth until 7 years postbirth, based on the child’s birth month. Given this, at the last time point (time=11), a child would be 7 years old. Time postbirth and assessed time since the birth of the child (coded as 0 before the birth of the child, up to a maximum value of 7). Time postbirth assessed whether the slope in the outcome variable changed after the birth of the child. It also enabled us to examine whether this divergence differed for mothers of CHPs compared with mothers of non-CHPs. Year before birth and year after birth examined possible changes in certain maternal health indicators (eg, number of physician visits) expected specifically during these periods.
Additional covariates included:
- receipt of an MSP health premium subsidy (yes/no), provided to those with low income, as indicated in the MSP file;
- an indicator of living in a low-income neighborhood (yes/no), defined as living in a census subdivision whose average household income is in the lowest provincial income quintile;
- maternal age at birth of the child, centered on the group mean of 26 years and divided by 10 to match the scale of other variables in the model;
- an indicator of whether the mother had another child with health problems born to the mother between 1996 and 2000 (yes/no).
We examined outcome and covariate descriptive statistics (proportions, means) separately for the 2 groups of mothers (CHPs, non-CHPs) at baseline (4 y before the birth of the child). With large sample sizes, even very small differences emerge as statistically significant. Therefore, effect sizes were computed using Cohen’s d to quantify our primary group comparisons using established criteria for small (0.2–0.5), medium (0.5–0.8), large (0.8–1.2), and very large (>1.2) effect sizes based on the standardized mean or percentage difference.27,28 Betas and SEs describe effect sizes from the regression models.
We examined bivariate correlations at baseline to look for collinearity among factors; we found only low to moderate levels of correlations. We used linear discontinuous (piecewise) growth-curve models to examine the number of maternal physician visits and the number of prescription drugs over time, and logistic discontinuous growth-curve models to examine the presence of selected maternal health conditions. Poisson models were explored for the count outcomes, but gave similar results to linear models: for ease of interpretation, we report linear model results.
Growth curve models have many applications, but generally refer to “ … statistical methods that allow for the estimation of inter-individual variability in intra-individual patterns of change over time.”29 In this case, we use a multilevel discontinuous (piecewise) approach to growth curve modeling.30 At level 1, predictors (ie, those predicting each individual time point of data) included temporal variables (time, time postbirth, year before birth, year after birth), receipt of a premium subsidy, and living in a low-income neighborhood to account for socioeconomic differences associated with the outcomes. Receipt of a premium subsidy and living in a low-income neighborhood were included at level 1 because they were potentially time-varying. Level 2 covariates, which can predict either the intercept or the slope of the level 1 variables, included CHP/non-CHP status of mothers, maternal age, and presence of another CHP child born between 1996 and 2000. A more detailed description of the models is presented in Appendix C (Supplemental Digital Content 1, http://links.lww.com/MLR/B752).
The Quasi-Newton optimization technique was used to fit the logistic model, selected because of its usability with large samples. Other optimization techniques required too much memory to compute. We conducted all analyses in SAS version 9.4.31
Table 1 compares the demographic and health characteristics of mothers of non-CHPs (n=20,282) and mothers of CHPs at baseline (4 y before birth of cohort child; n=1351). The 2 groups of mothers did not differ on age, percentage receiving a premium subsidy, or percentage living in a low-income neighborhood at baseline (Cohen’s d<0.2). However, mothers of CHPs were more likely to have another CHP born between 1996 and 2000 compared with mothers of non-CHPs (Cohen’s d=0.34), and had a higher number of physician visits (Cohen’s d=0.32) and more types of prescribed medication (Cohen’s d=0.35). There was very little difference between the groups at baseline on presence of selected physical health conditions or acute-care hospitalization (Cohen’s d<0.2).
Figure 1 (see also Table 2 for more detail) describes changes in the number of physician visits per year for the 2 groups of mothers in the presence of all temporal predictors and covariates that were statistically significant in the final model. Four years before the child’s birth, mothers of CHPs had a greater number of physician visits than did mothers of non-CHPs [Comparison A: beta (SE)=2.98 (0.21)]. Over time, the number of physician visits declined for all mothers [beta (SE)=−0.61 (0.02)], and this decline was similar for both groups of mothers [Comparison B: beta (SE)=−0.14 (0.08)]. As would be expected, mothers in both groups showed an increase in the number of physician visits in the year before the child’s birth [Comparison C: beta (SE)=1.46 (0.06)], an effect that was more pronounced for mothers of CHPs [Comparison D: beta (SE)=0.62 (0.23)]. There was also an increase in the number of physician visits in the year after the child’s birth for all mothers [Comparison E: beta (SE)=2.08 (0.06)], an effect that did not differ between groups [Comparison F: beta (SE)=0.26 (0.21)]. In the 6-year period from the child’s first to seventh birthdays, the decline in the number of physician visits slowed [Comparison G: beta (SE)=0.53 (0.03)], such that the number of physician visits in a year was stable for mothers of non-CHPs postbirth, while mothers of CHPs experienced a net increase in their annual number of physician visits during this time [Comparison H: beta (SE)=0.41 (0.11)].
Figure 2 shows largely similar changes in the number of types of medication per year for the 2 groups of mothers. Four years before the child’s birth, mothers of CHPs were prescribed more types of medication than were mothers of non-CHPs [Comparison A: beta (SE)=0.79 (0.06)]. Although there was a decline in the number of prescriptions over time [beta (SE)=−0.09 (0.01)], this decline was similar for both groups [Comparison B: beta (SE)=−0.02 (0.02)]. There were no important changes in number of types of medication in the year before birth, whereas there was an increase in the number of types of medication in the year after the child’s birth for all mothers [Comparison E: beta (SE)=0.31 (0.01)]: this increase did not differ between groups [Comparison F: beta (SE)=−0.02 (0.06)]. In the 6-year period from the child’s first to seventh birthdays, an initial decline in the number of medication types prescribed over time was offset by a subsequent increase in the number of types of medication prescribed [Comparison G: beta (SE)=0.18 (0.01)], but mothers of CHPs experienced a steeper increase than mothers of non-CHPs [Comparison H: beta (SE)=0.12 (0.03)]. These increases translate to a predicted mean number of types of medication from 1.8 to 2.3 for mothers of healthy children and from 2.8 to 3.7 for mothers of CHPs at 2 and 7 years postbirth, respectively.
Figure 3 describes changes in the predicted proportion of mothers diagnosed with a one or more of our selected physical health conditions. At baseline (4 y before birth), a greater percentage of mothers of CHPs were diagnosed with a selected health condition compared with mothers of non-CHPs [Comparison A: 42.5% vs. 30.9%; odds ratio (OR)=1.54]. The probability of being diagnosed with a selected health condition declined with time (OR=0.85), though this difference did not differ for mothers with or without CHPs [Comparison B: OR=1.0]. All mothers showed an increased likelihood of diagnosis of a selected health condition in the year before birth [Comparison C: OR=1.09], an increase that was significantly higher for mothers of CHPs [Comparison D: OR=2.10 for mothers of CHPs in year before birth compared with OR=1.54 for mothers of CHPs in other years, OR=1.09 for mothers of non-CHPs in year before birth, and OR=1.00 (ref) for mothers of non-CHPs in other years]. The likelihood of being diagnosed with a selected health condition also increased in the year after birth (Comparison E: OR=1.31), an increase that did not differ between groups (Comparison F: OR=1.06). In the period from the child’s first to seventh birthdays, there was an increase in the risk of being diagnosed with a selected health condition (Comparison G: OR=1.19), but the increase did not differ between the 2 groups (Comparison H: OR=1.06).
This study is the first to explore changes in maternal health in the years before, during, and after the birth of a child with a health problem, using linked population administrative health data. Our findings suggest that mothers of CHPs exhibit more health problems than mothers of non-CHPs before the birth of their child, show a steeper spike at the perinatal period, and a steeper slope of increasing health problems at 7 years postbirth for some outcomes.
Mothers of CHPs exhibit more service use, more types of medications prescribed, and a greater likelihood of having a selected health condition than non-CHP mothers well before the birth of the child. While a higher proportion of CHP mothers had a second, older child with health problems, our findings suggest that, even after accounting for this, the pattern of poorer health among mothers with CHPs remained. Given that health differences between mothers of CHPs and non-CHPs predate the child’s birth, our findings argue against a simple stress process model that attributes such maternal health effects solely to the strains of caregiving. Determining whether such preexisting differences between groups are primarily genetic, environmental, or mediated by other variables requires further investigation.
The 7-year period postbirth further differentiated the two groups of mothers. Mothers of CHPs showed more physician visits and more types of prescribed medications (but were not more likely to have one or more of our selected health conditions) compared with non-CHP mothers. In contrast, mothers of non-CHPs showed shallower increases in types of medication and selected health conditions, and even a slight decrease in physician visits over the same period. The diverging patterns, coupled with the change in trajectory from the prenatal period, is consistent with the notion that challenges associated with caring for a child with health problems may serve as an independent predictor of maternal health.
This work demonstrates how population-based health administrative data can be used to explore challenging, policy relevant questions such as how maternal health changes over time before and after specific landmark life events. Our approach has several advantages over previous studies. Previous large-scale surveys have been analyzed to demonstrate that child health can impact maternal health over many years,10,32 but such survey-based approaches are typically limited in the fidelity with which changes over time can be examined, and often fail to include clear comparison groups. They also typically measure outcomes at specific, predetermined data collection periods, rather than at times tied to landmarks such as the birth of a child. The power of these large, population-based datasets, the variety of outcomes measured over time, the ability to link to specific “landmark” events, and the potential for inclusion of a comparison, makes our approach particularly useful to those interested in exploring how other life events (eg, births, deaths, puberty, youth transitions to adulthood, etc.) impact child and maternal health.
This work has several limitations that warrant consideration. First, the data employed were relatively old (children born in 2000, outcomes obtained until 2007), a fact necessitated partly by our extensive follow-up period, and partly by difficulties and delays in data access and extraction. Replication with more recent data would be worthwhile. Second, correlations between children’s and mother’s use of health services might have been introduced by family physicians who treat both child and mother; in such cases, increased service use may reflect physician practice (or parental help-seeking) rather than poorer health. Physician identifiers were not available for this analysis. Third, we did not include the overall number of children in the family (including those who were born after 2000) because this information was not available. Fourth, the results are based on a single, 1-year cohort (children born in 2000). While we have no reason to believe the effects described are specific to this cohort, it is possible that some changes over time reflect secular trends rather than changes due to the caregiving situation. Fifth, our report does not include any maternal psychological outcomes; we attempted to include one (mood/anxiety disorders), but the various linear and nonlinear models we examined with the available data would not converge likely due to model misspecification. More specifically, the specified models had a poor fit to the data due to the high variance in the pattern of change in mood/anxiety over time. It is possible that model misspecification occurred because of under-reporting of counseling services in the administrative data as these services are not covered by the MSP. Sixth, we note that only mothers in British Columbia continuously registered with the MSP were included. As such, we cannot be certain that these results generalize to all mothers in British Columbia. Seventh, our use of heterogenous groupings of maternal and child health outcomes limits our ability to link these effects to specific caregiving situations or child conditions; more work on specific subgroups is a clear next step. Eighth, due to the nature of administrative data, we were limited in terms of the demographic variables we could include (eg, using neighborhood income quintile vs. family income) as well as indicators of severity of illness. Finally, we note that the health status of children in our cohort can change over time, with some children in the non-CHP group being identified with future health problems, and perhaps some in the CHP group outgrowing their health problems. Such effects may serve to reduce reported effect sizes.
This study shows that mothers of CHPs not only experience more health problems up to 4 years before the child’s birth, but that their health diverges from mothers of non-CHPs in the 7-year period postbirth. The health challenges of mothers of CHPs may therefore stem both from preexisting maternal conditions as well as caring for a child with a health problem. Interventions need to address existing maternal health issues, but preparation for challenges in the period of time around and after birth may be a particularly useful way to ameliorate or prevent health issues among mothers of children with a health problem.
The authors would like to thank Kelly Carroll for help with preparation of the manuscript, and Population Data British Columbia and the Data Stewards from the BC Ministry of Health for enabling extraction and analysis of the data.
1. Goudie A, Narcisse MR, Hall DE, et al. Financial and psychological stressors associated with caring for children with disability. Fam Syst Health. 2014;32:280–290.
2. Brehaut JC, Kohen DE, Raina P, et al. The health of primary caregivers of children with cerebral palsy: how does it compare with that of other Canadian caregivers? Pediatrics. 2004;114:e182–e191.
3. Breslau N, Staruch KS, Mortimer EA. Psychological distress in mothers of disabled children. Am J Dis Child. 1982;136:682–686.
4. Smyth-Staruch K, Breslau N, Weitzman M, et al. Use of health services by chronically ill and disabled children. Med Care. 1984;22:310–328.
5. Lach LM, Kohen D, Garner RE, et al. The health and psychosocial functioning of caregivers of children with neurodevelopmental disorders. Disabil Rehabil. 2009;31:607–618.
6. Brehaut JC, Kohen DE, Garner RE, et al. Health among caregivers of children with health problems: findings from a Canadian population-based study. Am J Public Health. 2009;99:1254–1262.
7. Cohen E, Horvath-Puho E, Ray JG, et al. Association between the birth of an infant with major congenital anomalies and subsequent risk of mortality in their mothers. JAMA. 2016;316:2515–2524.
8. Kohen DE, Brehaut JC, Garner RE, et al. Conceptualizing childhood health problems using survey data: a comparison of key indicators. BMC Pediatr. 2007;7:40–53.
9. Klassen AF, Raina P, McIntosh C, et al. Parents of children with cancer: Which factors explain differences in health-related quality of life. Int J Cancer. 2011;129:1190–1198.
10. Brehaut JC, Garner RE, Miller AR, et al. Changes over time in the health of caregivers of children with health problems: growth-curve findings from a 10-year canadian population-based study. Am J Public Health. 2011;101:2308–2316.
11. Goode KT, Haley WE, Roth DL, et al. Predicting longitudinal changes in caregiver physical and mental health: a stress process model. Health Psychol. 1998;17:190–198.
12. Schulz RH, Williamson GM. A 2-year longitudinal study of depression among Alzheimer’s caregivers. Psychol Aging. 1991;6:569–578.
13. Cohen CA, Colantonio A, Vernich L. Positive aspects of caregiving: rounding out the caregiver experience. Int J Geriatr Psychiatry. 2002;17:184–188.
14. Arim RG, Kohen DE, Brehaut JC, et al. Developing a non-categorical measure of child health using administrative data. Health Rep. 2015;26:9–16.
15. Arim RG, Guevremont A, Kohen D, et al. Exploring the Johns Hopkins Aggregated Diagnosis Groups in administrative data as a measure of child health. Int J Child Health Hum Dev. 2017;10:19–29.
16. Centers for Disease Control and Prevention. Mental health surveillance among children-United States, 2005-2011. MMWR Suppl. 2013;62:1–35.
17. British Columbia Ministry of Health. Medical Services Plan (MSP) Payment Information File. Population Data BC. Data Extract. MOH. 2012. Available at: www.popdata.bc.ca/data
. Accessed March 4, 2019.
18. Canadian Institute for Health Information. Discharge Abstract Database (Hospital Separations). Population Data BC. Data Extract. MOH. 2012. Available at: www.popdata.bc.ca/data
. Accessed March 4, 2019.
19. British Columbia Ministry of Health. PharmaNet. BC Ministry of Health. Data Extract. Data Stewardship Committee. 2012. Available at: www.popdata.bc.ca/data
. Accessed March 4, 2019.
20. British Columbia Ministry of Health. Consolidation File (MSP Registration & Premium Billing). Population Data BC. Data Extract. MOH. 2012. Available at: www.popdata.bc.ca/data
. Accessed March 4, 2019.
21. The Johns Hopkins Bloomberg School of Public Health. The Johns Hopkins ACG System: Technical Reference Guide Version 10.0. 2011.
22. Bethell CD, Read D, Blumberg SJ, et al. What is the prevalence of children with special health care needs? Toward an understanding of variations in findings and methods across three national surveys. Matern Child Health J. 2008;12:1–14.
23. Hinds AM, Bechtel B, Distasio J, et al. Changes in healthcare use among individuals who move into public housing: a population-based investigation. BMC Health Serv Res. 2018;18:411–424.
24. Lix L, Yogendran M, Burchill C, et al. Defining and validating chronic diseases: an administrative data approach. Faculty of Medicine, University of Manitoba, Winnipeg MB, Manitoba Centre for Health Policy. 2006.
25. Statistics Canada. National Population Health Survey. Ottawa, ON, Canada: Statistics Canada; 1996.
26. Statistics Canada. National Longitudinal Study of Children and Youth. Ottawa, ON, Canada: Statistics Canada; 1996.
27. Cohen J. Statistical Power Analysis for the Behavioral sciences, 2nd ed. Hillsdale, NJ: Lawrence Earlbaum Associates; 1988.
28. Rosenthal JA. Qualitative descriptors of strengh of association and effect size. J Soc Serv Res. 1996;21:37–59.
29. Curran PJ, Obeidat K, Losardo D. Twelve frequently asked questions about growth curve modeling. J Cogn Dev. 2010;11:121–136.
30. Singer J, Willett J. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York, NY: Oxford University Press; 2003.
31. SAS Institute. 2016. Sixth Edition. Cary, NC.
32. Smith AM, Grzywacz JG. Health and well-being in midlife parents of children with special health needs. Fam Syst Health. 2014;32:303–312.
caregiver health; maternal healthchild health problems; administrative health data; longitudinal analysis
Supplemental Digital Content
Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.