OBJECTIVE. The objective of this study was to determine the association between parental depression and pediatric health care use patterns.
METHODS. We selected all children who were 0 to 17 years of age, enrolled in Kaiser Permanente of Colorado during the study period July 1997 to December 2002, and linked to at least 1 parent/subscriber who was enrolled for at least 6 months during that period. Unexposed children were selected from a pool of children whose parents did not have a depression diagnosis. Outcome measures were derived from the child's payment files and electronic medical charts and included 5 categories of use: well-child-care visits, sick visits to primary care departments, specialty clinic visits, emergency department visits, and inpatient visits. We compared the rate of use per enrollment month for these 5 categories between exposed and unexposed children within each of the 5 age strata.
RESULTS. Our study population had 24391 exposed and 45274 age-matched, unexposed children. For the outcome of well-child-care visits, teenagers showed decreased rates of visits among exposed children. The rate of specialty department visits was higher in exposed children in the 4 oldest age groups. The rates of both emergency department visits and sick visits to primary care departments were higher for exposed children across all 5 age categories. The rate of inpatient visits was higher among exposed children in 2 of the 5 age groups.
CONCLUSIONS. Overall, having at least 1 depressed parent is associated with greater rate of emergency department and sick visits across all age groups, greater use of inpatient and specialty services in some age groups, and a lower rate of well-child-care visits among 13- to 17-year-olds. This pattern of increased use of expensive resources and decreased use of preventive services represents one of the hidden costs of adult depression.
- health services research
- health care provider/services
- parental influence
- maternal mental health
Numerous studies have shown a high prevalence of depression among parents, ranging as high as 47% in some pediatric settings.1–6 Prevalence is highest in parents who care for a chronically ill family member.7–9 Studies of the impact of parental depression on child health outcomes have found positive associations with adverse child behavioral,10–19 developmental,20–23 psychological,14,24–28 and physiologic outcomes.29
Studies that have examined the association between parental depression and child health care use outcomes have been inconsistent in their findings. Two studies found lower rates of well-child-care (WCC) visits among children of depressed parents31; other studies found no association with WCC indicators.32–35 Studies have found an association between parental depression and child acute-care use indicators, including hospitalization,29,32,34,36,37 and emergency department (ED) visits.35,36,38 Only 2 studies found no association between a parent's depression and ED visits.33,37
Most of these previous studies were limited by small sample sizes, by studying only 1 type of use, by studying use in only 1 type of setting (eg, clinic, nursery), or by relying on the parent's report of use to measure outcomes. No previous studies combined a wide array of use outcomes in the context of a large sample size, and none did this in the context of a closed-model health maintenance organization (HMO) setting, which allows nearly total capture of all health care use.
Closing these literature gaps has important implications for families and their health care providers. By strengthening and broadening our understanding of the association between parental depression and child health care outcomes, we can improve clinicians' assessment and management of the pediatric patient in the context of the child–parent dyad. By demonstrating the magnitude of this association, we provide additional impetus to improve mental health services for parents and help policy-makers understand more about the hidden costs of adult depression. The objective of this study was to assess the association between parental depression and pediatric health care use. We hypothesized that children of depressed parents would have lower use of preventive care and higher use of emergent services than children of nondepressed parents. This association was hypothesized to be present for both maternal and paternal depression and across all age strata of children.
The data for this study were drawn from the Kaiser Permanente of Colorado (KPCO) membership system. The study period for these analyses was July 1997 to December 2002. Our study design was a retrospective, matched-cohort design.
We identified all children who were 0 to 17 years of age and enrollees of KPCO during the study period July 1997 to December 2002. Children were linked to “parents” by identifying subscribers through whom the child received eligible care. Over time, many children had >1 subscriber. We considered any subscriber who was linked to a child's record to be a “parent.” This method missed parents who were not primary subscribers to KPCO, such as spouses who consistently were dependents during the entire study period, and may have included nonparents, including grandparents or unrelated subscribers. We included only those children who were linked to at least 1 parent/subscriber who had been a member for at least 6 continuous months during this time. Only 1 child was randomly selected from each family within each predefined age group (3–11 months, 1–2 years, 3–5 years, 6–12 years, and 13–17 years).
Children who had any parent with a depression diagnosis were classified as exposed for these analyses. Parents' diagnoses were obtained from the visit records and outside claim records. A parent was considered to have depression when their records contained any of the International Classification of Diseases, Ninth Revision (ICD-9) depression codes in Table 1.
Unexposed children were selected from a pool of children whose parents did not have a depression diagnosis and whose parents also did not have any other mental health condition or prescription antidepressant use. Other mental health conditions were identified using the ICD-9 codes between 290 and 316. Excluded codes were the depression codes listed in Table 1 and codes that commonly are used in primary care for non–mental health issues: 302.7 psychosexual dysfunction (impotence), 305 nondependent drug use (includes tobacco use), 307.81 tension headache, and 315 developmental disorders. Pharmacy records provided information on prescription antidepressant use.
For each child, we defined an index date indicating the start of the period of captured use data. For exposed children, we defined this starting point as the latter of 2 dates: (1) the first captured parental depression date or (2) the 90th day after the child's KPCO enrollment date. Children were required to be members for at least 3 months before the index date to avoid capturing use during initial months when a new member's use often is inflated. This matching requirement resulted in exclusion of all children who were 0 to 3 months from our sample. We excluded children whose membership duration was <1 month after the index date; this was a relatively rare finding (<3%). A total of 24413 exposed children met these eligibility criteria.
Up to 2 unexposed children were matched to each exposed child by age and membership eligibility criteria. We chose to match by age because of the great age-related variability in health care use, fostered, in part, by nationally endorsed childhood immunization and WCC visit schedules. We chose to match by enrollment period to align the timing of use data capture for exposed and unexposed children, to minimize problems that changes in administrative data elements over time might introduce. First, we selected unexposed children with birth dates within a 2-week window of the exposed child's birth date. We then set the index dates for the unexposed children to the date for the matched exposed child and applied the same membership eligibility minimum requirements of 3 months before and 1 month after the index date.
Of the 24413 exposed children, 24391 were matched to at least 1 unexposed child by age and membership eligibility criteria. Two control matches were found for 85.6% (20883). This produced a total N of 69665 for these analyses.
Outcome measures were derived from the KPCO use database and included 5 categories of use: WCC visits, sick visits to primary care departments, specialty clinic visits, ED visits, and inpatient visits. Primary care department providers include Family medicine, internal medicine, pediatrics, and obstetrics/gynecology practitioners. All visits to other departments were classified as “specialty” visits.
We included a risk adjustment variable to control for possible morbidity differences between exposed and unexposed children. We used a pharmacy-based disease score that was developed by Fishman and Shay39 and applied this to pharmacy dispensings during the 12 months after the index date. The 33 disease indicator groups include both physical and mental health categories. We evaluated several different variables for possible morbidity adjustment: (1) the risk score (RxRisk) as it was originally developed and weighted to predict costs in the following year; (2) a summed count of the 33 disease indicators used in the score; (3) a summed count of the disease indicators excluding (a) sickle cell disease, which needed reprogramming to capture appropriately only penicillin use with concurrent folic acid use and which is low prevalence in our region, and (b) the liver disease indicator, which contained a drug that often is used to treat pediatric constipation at KPCO; and (4) selected individual disease indicators entered as separate variables. The count of diseases excluding the 2 questionable indicators (modified RxRisk count) provided the most consistent adjustment across the various use outcomes in this study and also decreased exposed versus unexposed differences to the greatest degree. We selected this morbidity adjustment as the best and most conservative adjustment for the analyses presented here.
Frequencies of categorical variables were contrasted by χ2 statistics. Wilcoxon rank sum tests were used for the chronic disease score and membership months variables because of the skewness of their distributions. We performed bivariate comparisons of characteristics of exposed and unexposed children within each age group.
We calculated a visit rate for each group by dividing the total number of each type of visit during the children's captured use period by the duration of that period. We compared rates of use for each of the 5 visit categories between exposed and unexposed children within each of the 5 age strata, yielding 25 comparisons. We estimated confidence intervals for these person-time rates using Fisher's exact method in PEPI software version 4.0.40 These rate contrasts provide magnitudes of visit differences between exposed and unexposed children while accounting for varying enrollment periods in the denominator calculations but do not adjust for matching.
Incidence rate ratios were estimated using conditional Poisson models. These models used counts of visits over time as the primary outcome, adjusted for varying enrollments by an offset variable and additionally accounted for matching with a stratum identifier. Both ratios of the descriptive rates and the incidence rate ratios from univariate conditional Poisson models were presented. Adjusted incidence rate ratios also are presented controlling for the child's gender, number of parents, and the pediatric RxRisk risk-adjustment variable described.
We estimated the excess visit rate that was attributable to having depressed parent(s) by multiplying the adjusted rate ratio by the unadjusted visit rate for unexposed children. For example, for infants, the exposed versus unexposed sick visit adjusted rate ratio is 1.14. This suggests that exposed infants have 14% more sick visits than unexposed infants, so the excess would be 0.14 × 487.1 (the unadjusted rate of sick visits for unexposed infants), which equals 68.2 visits per 100 person-years.
Analyses were completed using SAS 9.1.41 This study was approved by the institutional review boards of the Children's Hospital of Denver and KPCO.
Our study population had 24391 exposed and 45274 unexposed children. Table 2 shows the comparison of gender, number of linked parents, number of chronic diseases, and membership duration between exposed and unexposed children within each age group. There were slightly more girls in the unexposed group, although this difference was not significant. The number of parents who were linked to the child's KPCO record was higher among exposed than unexposed children. Exposed children had longer periods of enrollment after the index date. The differing enrollment times are accounted for in the Poisson models. As expected, exposed children had a higher mean number of chronic diseases.8
Of the exposed children, 88.2% had 1 depressed parent, 11.7% had 2, and 0.1% had 3. Gender of depressed parents was 79.7% female and 31.7% male; these figures include the 11.4% of children with depressed parents of both genders. Of all depressed parents, 83.9% were on antidepressant medications.
Unadjusted use rates are presented in Table 3, which presents the visit rate for exposed and unexposed children. Sick visits to primary care sites were the most common visit type across all age groups, even in the infant age category, when frequent WCC visits are the standard of care. ED visit rates were higher than specialty visit numbers in the youngest ages, but this pattern reversed among older children.
The most common departments in the specialty visits, by percentage of all subspecialty visits, were mental health (23.9%), optometry (18.6%), orthopedics (10.7%), head and neck surgery (9.6%), ophthalmology (6.6%), dermatology (5.6%), and allergy (5.5%). Mental health, optometry, and orthopedics visits increased with increasing age, with >80% of visits in each of these 3 categories among 6- to 17-year-olds.
Differences in visit counts for exposed and unexposed children were apparent for several types of visits in these unadjusted numbers (Table 3). WCC visit rates for exposed children were either similar to or slightly lower than those of unexposed children, although none of these differences was significant. In contrast, exposed children had higher rates of sick visits and ED visits in all ages. Univariate rates of specialty visits also generally were higher among exposed children, although confidence intervals were nonoverlapping only for children who were 3 years or older. Hospital rates were higher for exposed children who were 6 years and older.
Table 4 improves the statistical comparison of exposed and unexposed children through the use of conditional Poisson models to add control of matching as well as adjusting for possible confounders, including gender, number of parents, and disease history counts. In Table 4, univariate rate ratios showed patterns of significance and direction that generally mirrored those seen in Table 3. Adding adjustment for possible confounders generally lowered the exposed versus unexposed differences primarily as a result of adjustment for the higher disease counts among exposed children.
For the outcome of WCC visits, our adjusted model showed a decreased rate of WCC visits among exposed children in the teenage category. The rates of both ED visits and sick visits to primary care departments were higher for exposed children in the adjusted models across all 5 age categories, and the rate of specialty department visits was higher in exposed children in the 4 oldest age groups. The rate of inpatient visits was higher among exposed children in the 2 oldest age groups.
Table 5 shows the excess visit rate, in visits per 100 person-years, that was attributable to having a depressed parent. The difference is greatest for sick visits to primary care departments, particularly in the youngest age group, in which exposed infants had 68 more visits per 100 person-years than did unexposed infants. For both sick visits and ED visits, the age-related trend was similar, with greater excesses in the youngest age groups. In contrast, the greatest excesses in specialty visit rates were among the older age groups.
In this study, children with at least 1 depressed parent had higher use of costlier forms of health care—using more ED, sick visit, specialty department, and inpatient services—than did children of parents without depression. Findings from our study, in general, support previous studies, although our study strengthens the evidence for these findings, because no previous studies combined a wide array of use outcomes in the context of a large sample size, and none did this in the context of a closed-model HMO setting. Ours also is only the second study to consider paternal as well as maternal depression as a predictor of health care use.
Our finding of no association between parental depression and a child's WCC visits in the 4 youngest age groups is consistent with previous findings. However, our finding that 13- to 17-year-olds had fewer WCC visits when they had a depressed parent is the first report that we could find of a significant association between parental depression and WCC visit rates in adolescents. This does not contradict previous studies, because all 4 previous analyses of the association between parental depression and WCC visits studied infants32,34,35 or preschool children only.33
Our finding of an association between parental depression and higher rates of sick visits to the primary care site is consistent with the 1 previous study of this association, authored by Mandl et al.35 That study found that women were twice as likely to exhibit depressive symptoms when their infant had >1 sick visit. We found no comparison studies that assessed the association between specialty visits and parental depression.
Regarding ED visits, previous literature is mixed, with some studies showing increased ED visits among children of depressed parents36 and others showing no association.33,37 By showing a strong association between ED visits and parental depression across all age groups and in a large patient population, our study greatly strengthens the evidence for this association.
The literature is more consistent regarding the association between parental depression and the child's likelihood of inpatient visit, with previous studies finding an odds ratio of 1.5 to 3.0.32,34,36–38 Our study found an association only in the 2 oldest age groups. Of the previous 5 analyses, 3 studied children who were younger than 2 years.32,34,36 Our finding of no association in the 3 youngest age groups is inconsistent with these previous findings and may be related to selection bias: the 2 US studies selected their subjects from families who attended primary care clinics, which may select for a population with a greater tendency to use services.32,36 Of the 2 previous analyses that involved older children, 1 studied 4- to 9-year-old children with asthma,38 and the other studied 6- to 23-year-olds; the latter study found higher rates of surgery-related inpatient visit among children of depressed parents but no difference in the rate of nonsurgical inpatient visits.37
The finding of higher excess visit rates for ED and sick visits among the youngest age groups may reflect the greater difficulty that depressed parents encounter confronting the challenge that is presented by figuring out the anatomic site and the severity of distress among preverbal children. It also may reflect the influence of ongoing postpartum depression, which can last up to 1 year after delivery. Because we used the adjusted rate ratio in calculating excess visits, we accounted for the effect of chronic illness; this is corroborated further by the finding that no excess inpatient visits were noted in these youngest age groups despite their higher excesses of emergent and urgent visit rates.
The clinical significance of the excess bed days depends to some degree on perspective. None of our rates exceeded 1 extra visit per patient per year, so the average impact of excess visits on families may not be clinically significant. However, even the smallest significant excess in visits—an excess of 1.3 inpatient visits per 100 patients in the 13- to 17-year age group—is significant from a health care financing standpoint.
The limitations of this study are related to limitations that are inherent in using medical chart databases. Several factors limit the accuracy of medical charts for disease identification in general, including incomplete or erroneous charts submitted by providers and limited clinical detail in the ICD-9 system. The use of administrative data for depression identification in particular presents additional challenges.42 Social stigma may discourage individuals from reporting mental health diagnoses, and suboptimal health care screening practices by providers may limit further individuals' ability to know and report mental illness.43,44 These limitations would result in misclassification of exposed children as unexposed children and would be expected to dilute the effect reported. The identification of “parents” by subscriber match also led to inexact linking of parents and children. This, too, likely would result in misclassification of exposed children as unexposed, which would bias our results to the null. Another set of limitations are related to dissimilarities between the exposed and unexposed groups. We found a higher number of parents linked to the child's KPCO record among exposed children than unexposed. This finding may result, in part, from the fact that having more parents gives a child more opportunity to have at least 1 depressed parent. We also found longer enrollment periods among exposed children. This may reflect a phenomenon that is similar to the effect of having more parents linked to one's file; namely, children with longer enrollments had more opportunity to have a parent receive a diagnosis of depression. Finally, because this is a retrospective analysis, we cannot make assumptions about causality. Despite adjusting for chronic illness, we cannot assume that causality starts with the parent's depression; perhaps it is the child's genuine need for more visits that has contributed to the parent's depression. In either causality scenario, efforts that are directed at improving the parent's depression are likely to be associated with improved outcomes for the parent–child dyad.
This study confirms previously reported associations in a large, closed-model HMO setting. The costlier patterns of health care use that are associated with parental depression raise important issues both for pediatric health care providers and for health care policy-makers. For pediatric providers, our study supports the conclusions of previous reports, which have called for increased screening for and treatment of depression in the parents of pediatric patients.45 Conducting routine, brief, maternal depression screening during WCC visits has been found to be feasible, successful, and well accepted and has resulted in specific pediatrician actions.46 For policy-makers, our findings suggest that interventions that are directed at improving rates of detection and treatment of parental depression will result in less costly health care use patterns for children of depressed parents.
This study was funded by grants D14HP00153 and D54HP00054 from the Health Resources and Services Administration and grant G06807 from the Children's Hospital Research Institute.
- Accepted October 20, 2006.
- Address correspondence to Marion R. Sills, MD, MPH, University of Colorado Health Sciences Center–Pediatrics, Children's Hospital of Denver, 1056 E 19th Ave B251, Denver, CO 80218. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Olson AL, DiBrigida LA. Depressive symptoms and work role satisfaction in mothers of toddlers. Pediatrics.1994;94 :363– 367
- ↵Kemper KJ. Self-administered questionnaire for structured psychosocial screening in pediatrics. Pediatrics.1992;89 :433– 436
- ↵Coghill SR, Caplan HL, Alexandra H, Robson K, Kumar R. Impact of maternal post-natal depression on cognitive development of young children. Br Med J.1986;292 :1165– 1167
- ↵Minkovitz CS, Strobino D, Scharfstein D, et al. Maternal depressive symptoms and children's receipt of health care in the first 3 years of life. Pediatrics.2005;115 :306– 314
- ↵Chung EK, CuCollum KF, Elo IT, Lee HJ, Culhane JF. Maternal depressive symptoms and infant health practices among low-income women. Pediatrics.2004;113 (6). Available at: www.pediatrics.org/cgi/content/full/113/6/e523
- ↵Weil CM, Wade SL, Bauman LJ, Lynn H, Mitchell H, Lavigne J. The relationship between psychosocial factors and asthma morbidity in inner-city children with asthma. Pediatrics.1999;104 :1274– 1280
- ↵Abramson JH, Gahlinger PM. Computer Programs for Epidemiologists: PEPI V. 4.0 [computer program]. Salt Lake City, UT: Sagebrush Press; 2001
- ↵SAS Institute. SAS/STAT 9.1 User's Guide. Cary, NC: SAS Institute; 2004
- ↵Lemelin J, Hotz S, Swensen R, Elmslie T. Depression in primary care. Why do we miss the diagnosis? Can Fam Physician.1994;40 :104– 108
- ↵Olson AL, Dietrich AJ, Prazar G, Hurley J. Brief maternal depression screening at well-child visits. Pediatrics.2006;118 :207– 216
- Copyright © 2007 by the American Academy of Pediatrics