Background And Objectives To understand the association between shared decision-making (SDM) and health care expenditures and use among children with special health care needs (CSHCN).
Methods We identified CSHCN <18 years in the 2002–2006 Medical Expenditure Panel Survey by using the CSHCN Screener. Outcomes included health care expenditures (total, out-of-pocket, office-based, inpatient, emergency department [ED], and prescription) and utilization (hospitalization, ED and office visit, and prescription rates). The main exposure was the pattern of SDM over the 2 study years (increasing, decreasing, or unchanged high or low). We assessed the impact of these patterns on the change in expenditures and utilization over the 2 study years.
Results Among 2858 subjects representing 12 million CSHCN, 15.9% had increasing, 15.2% decreasing, 51.9% unchanged high, and 17.0% unchanged low SDM. At baseline, mean per child total expenditures were $2131. Over the 2 study years, increasing SDM was associated with a decrease of $339 (95% confidence interval: $21, $660) in total health care costs. Rates of hospitalization and ED visits declined by 4.0 (0.1, 7.9) and 11.3 (4.3, 18.3) per 100 CSHCN, and office visits by 1.2 (0.3, 2.0) per child with increasing SDM. Relative to decreasing SDM, increasing SDM was associated with significantly lower total and out-of-pocket costs, and fewer office visits.
Conclusions We found that increasing SDM was associated with decreased utilization and expenditures for CSHCN. Prospective study is warranted to confirm if fostering SDM reduces the costs of caring for CSHCN for the health system and families.
- AHRQ —
- Agency for Healthcare Research and Quality
- CI —
- confidence interval
- CSHCN —
- children with special health care needs
- ED —
- emergency department
- IOM —
- Institute of Medicine
- IRB —
- institutional review board
- MEPS —
- Medical Expenditure Panel Survey
- NHIS —
- National Health Interview Survey
- SDM —
- shared decision-making
What’s Known On This Subject:
Children with special health care needs (CSHCN) account for more than one-third of pediatric health care costs. Little is known regarding the impact of shared decision-making (SDM) over time on child health care expenditures and utilization.
What This Study Adds:
In a national sample, we found that increasing SDM was associated with decreased health care costs and utilization for CSHCN. Results support prospective studies to determine if pediatric interventions to foster SDM reduce the financial burden of caring for CSHCN.
In shared decision-making (SDM), families and clinicians both participate in decisions, exchange information, express preferences, and negotiate the treatment plan.1 Given benefits of SDM in increasing patients’ knowledge, decreasing uncertainty, and limiting the overuse of treatments that patients do not value,2 the Institute of Medicine (IOM) recently encouraged research assessing the comparative effectiveness of SDM3 and the 2010 Patient Protection and Affordable Care Act supports the implementation of SDM in clinical settings.4 In pediatrics, partnership between families and clinicians is one of the Maternal and Child Health Bureau’s 6 core measures of the care of children with special health care needs (CSHCN)5 and a primary attribute of the medical home.6,7 Despite this emphasis, little is known regarding the impact of SDM on children’s health care costs and utilization.
CSHCN, those who have or are at increased risk of a chronic physical, developmental, behavioral, or emotional condition and require health and related services beyond those required by children generally,8 are an ideal population to study the impact of SDM on health care costs and utilization. CSHCN have health care expenses that are 3 times higher than other children.9 Although they represent approximately 15% of the population, CSHCN account for 33.6% or more of total health care costs attributable to children.9–11 CSHCN also have 4 times the number of hospitalizations, twice as many physician visits, 1.5 times as many emergency department (ED) visits, and receive 5 times the number of prescriptions.9 Reflecting the impact of higher health care utilization and costs, approximately 40% of families of CSHCN experience burdensome out-of-pocket expenses even when they are eligible for safety net programs.12,13
Prior literature has described health care costs and utilization for CSHCN, but has not examined the impact of SDM on these outcomes. We conducted a longitudinal analysis of a nationally representative dataset to address this knowledge gap. We hypothesized that both increases in SDM and higher absolute levels of SDM over time would lead to decreased costs and utilization of health services.
Study Design and Data Source
We conducted a longitudinal analysis of the Medical Expenditure Panel Survey (MEPS), administered annually by the Agency for Healthcare Research and Quality (AHRQ) and previously used to study expenditures for CSHCN.9,11 All children <18 years old were followed for 2 years. Between 12 810 and 14 828 households were sampled annually from the US civilian, noninstitutionalized population drawn from the previous year’s National Health Interview Survey (NHIS).14 The person from each household who was most knowledgeable about the health of its members provided information on health status, health insurance, and health care utilization. Household interviews were supplemented by surveys from medical providers, health insurers, pharmacies, and employers who provided additional health expenditure and utilization data.
The study sample included all children <18 years of age in MEPS panels 7 to 10 (2002–2006). From this sample, CSHCN were identified using the validated CSHCN Screener that identifies children based on functional limitations or health care needs.15,16 Children were excluded if they had no usual source of care or their household did not respond to any of the items used to create the SDM measure. MEPS response rates for completion of all survey rounds for the years considered ranged from 58.3% to 64.7%.14 We were able to generalize results to the general population of CSHCN in the United States by accounting for the stratification, clustering, and unequal probabilities of selection and response in the complex survey design.
We examined total health care expenditures as well as multiple components of cost. Total health care expenditures included direct payments to health care providers for outpatient visits, home health care, prescriptions, dental visits, hospital stays, ED visits, and other medical equipment and expenses.11,14 We separately considered out-of-pocket (copayments and payments not reimbursed by insurance),17 office-based, inpatient, ED, and prescription costs. To avoid bias from extreme outliers and violations of the assumptions of normality of our statistical models,9 we calculated all expenditure results with the top 2.5% of values trimmed to the 97.5 percentile. As a sensitivity analysis that also minimized bias from outliers, we assessed the cost percentile rank for each subject for each year for these outcomes.
We also assessed the impact of SDM on health care utilization, including the rates of hospitalizations and ED visits per 100 children per year, as well as office visits and prescriptions per child per year.
As we have previously reported in detail,18 we determined families’ participation in SDM using a latent class analysis of responses to 7 separate MEPS items that address the 4 components of SDM in the most widely accepted definition (Table 1).1 Latent class models identify homogeneous groups of people according to their observed response patterns.18–20 Items were drawn from the Access to Care and the communication and quality-oriented Consumer Assessment of Health Plans Survey sections of the MEPS survey.
For each study year, each child’s household was initially categorized to have low, intermediate, or high participation in SDM. Because the low SDM group represented <1% of the population, results were unchanged when the low and intermediate groups were combined. Then, to demonstrate the impact of both the level of and change in SDM over the 2 study years, we classified households into 4 patterns: increasing, decreasing, unchanged low (combines intermediate and low), or unchanged high SDM. Increasing and decreasing SDM were defined based on movement between the initial 3 SDM categories between the 2 study years. For example, a change from low to intermediate, low to high, or intermediate to high SDM was categorized as an increasing pattern. We determined that these subgroups had similar effects on study outcomes before combining them.
We considered additional variables that might confound the relationship between SDM, expenditures, and utilization. These covariates consisted of demographic characteristics including the child’s age (0–4 vs 5–12 vs 13–17 years), gender, race (white, black, other) and Hispanic ethnicity, region of residence (Northeast, Midwest, South, West), parental education (no high school diploma, high school diploma, bachelor’s degree, graduate-level degree, or other degree), and household income (poor, <100% of the applicable poverty line; near poor, 100% to <125%; low, 125% to <200%; middle, 200% to <400%; high, ≥400%), as well as any private health insurance (versus others). To ensure that findings were not confounded by child health, we adjusted for general health status based on the overall score (low, <15; medium, 15 to <20; and high, ≥20) from 5 Likert-scaled items derived from the Child Health Questionnaire, General Health Subscale.18,21 Specifically, we considered 5 patterns of health status: unchanged low, medium, or high, increasing, and decreasing.
Characteristics of the Study Sample
We initially described the weighted proportion of CSHCN in MEPS included versus excluded from the study and compared the characteristics of CSHCN with each pattern of SDM using χ2 tests. Following guidance from AHRQ, all statistical analyses were conducted using a statistical package (Stata) designed specifically for the analysis of longitudinal data from the weighted, clustered, and stratified MEPS survey.22 This approach provides conservative estimates accounting for repeated measures over time.22P values of <.05 were considered significant.
Assessing the Impact of SDM on Expenditure and Utilization Outcomes
To assess the impact of SDM on all expenditure outcomes in the presence of skewed cost data, we used 2 approaches that would be resistant to high-cost outliers. First, we calculated the mean cost for each child during each of the 2 study years, with the top 2.5% of values trimmed to the 97.5 percentile. Second, without trimming, we estimated the mean expenditure percentile rank for CSHCN with each of the observed patterns of SDM for each study year. We next calculated the change in expenditures as well as expenditure percentile rank with 95% confidence intervals (CIs) for those with increasing, decreasing, or unchanged high or low SDM.
We used the same approach for all utilization measures, first describing the mean rate of hospitalizations, ED visits, office visits, or prescriptions for each child in each study year by SDM pattern and then calculating the change in rates with 95% CIs for each of these patterns.
Linear regression was then used to estimate the impact of SDM on expenditures and utilization while adjusting for differences in children’s characteristics across the distinct SDM patterns. For all expenditure and utilization outcomes, linear regression models with robust variance estimates that reflect the complex survey design were used with the change in expenditures or utilization between study years as the outcome and pattern of SDM as the independent variable. Expenditure results were confirmed by models using cost percentile rank as the outcome.
We then constructed models with and without covariates. Since the inclusion of covariates did not change the association of SDM with expenditures or utilization, unadjusted results are presented for all analyses. Models were implemented in Stata 10 and 11 (StataCorp, College Station, TX). The Children’s Hospital of Philadelphia Institutional Review Board (IRB) determined this study to be IRB exempt.
The study sample, which included 2858 CSHCN, was representative of 12 million US children. This sample, composed of those with a pattern of SDM assigned and a usual source of care, included 90% of the weighted US population of CSHCN (Table 2). Those excluded were significantly more likely to be of black or other non-white race (P < .001), to be Hispanic (P = .03), to have lower levels of parental education (P = .02) and higher levels of poverty (P < .001), and to have no private health insurance (P < .001).
In the weighted study population, 15.9% had increasing, 51.9% unchanged high, 17.0% unchanged low, and 15.2% decreasing SDM (Table 2). Overall, these groups had similar demographic characteristics; however, children whose families reported increasing or unchanged high SDM were less likely to be 0 to 4 years of age (P = .009). Additionally, children whose families reported decreasing or unchanged high SDM were more likely to have private insurance (P = .04), and those whose families reported unchanged high SDM were more likely to have persistently high overall health (P = .0005).
Association of SDM With Health Care Expenditures
At baseline, there were no significant differences between those with distinct SDM patterns for any cost or utilization outcome. CSHCN had mean total, out-of-pocket, office-based, inpatient, ED, and prescription expenditures of $2832, $454, $645, $617, $82, and $624, respectively. With the top 2.5% of expenses trimmed, values were $2131, $389, $557, $127, $64, and $478 (Table 3). We present trimmed values for all subsequent results.
Between years 1 and 2 of the study, increasing SDM was associated with a significant decrease in total health care expenditures (−$339 (95% CI: −$660, −$21) (Table 2). Total costs did not change significantly over time for those with any other SDM pattern. During the study period, prescription expenditures increased for all groups.
When we assessed the relative differences in the change in costs from year 1 to 2 between those with each SDM pattern (Table 4), we found that those with increasing SDM had significantly lower total and out-of-pocket health care expenditures compared with those with decreasing SDM with relative differences of −$584 (−$1131, −$38) and −$142 (−$265, −$19). No significant contrasts were observed between the unchanged high and low SDM groups for any outcome. These expenditure results were confirmed in secondary analyses with the cost percentile rank as the outcome.
Association of SDM With Health Care Utilization
At baseline, CSHCN had rates of 6.5 hospitalizations and 27.0 ED visits per 100 children per year and 6.1 office visits and 7.8 prescriptions per child per year.
We found that increasing SDM was associated with significant decreases in hospitalizations, ED visits, and office visits over time (Table 3). Rates of hospitalizations and ED visits for CSHCN with increasing SDM declined by 4.0 (0.1, 7.9) and 11.3 (4.3, 18.3) per 100 children per year, respectively. Results showed a mean drop of 1.2 (0.3, 2.0) outpatient visits per child per year with increasing SDM. When we examined relative differences over time between those with each SDM pattern, we found that those with increasing SDM had 1.7 (0.3, 3.2) fewer office visits per child per year compared with those with decreasing SDM, 1.3 (0.3, 2.3) fewer visits compared with those with unchanged high SDM, and 2.2 (0.2, 4.2) fewer visits compared with those with unchanged low SDM (Table 5). Over time, those with increasing SDM had 4.4 (0.1, 8.7) fewer hospital visits and 10.3 (1.9, 18.6) fewer ED visits per 100 CSHCN per year than those with unchanged high SDM. We observed no differences over time in prescription rates. As with expenditure outcomes, we found no relative differences in utilization over time when comparing those with unchanged high versus low SDM.
In a US sample representing 12 million CSHCN, we found that increasing SDM was associated with decreases in health care expenditures and utilization. For those with increasing SDM, total costs declined by $339. Relative to those with decreasing SDM, CSHCN with increasing SDM had $584 lower total costs over the study period, and costs trended lower for those with increasing SDM in all expense categories studied except for prescriptions. Overall, savings were substantial compared with the mean baseline costs of $2131 across all CSHCN. These results, confirmed in sensitivity analyses and unchanged in analyses that considered multiple covariates, including trends in general health status, are consistent with our hypothesis that increases in SDM over time would result in overall savings.
Although increasing SDM was associated with lower costs, our results did not show that higher absolute levels of SDM resulted in savings. No significant absolute or relative change in total costs was observed over time among those with unchanged high or low SDM. In the context of decreases in total expenditures with increasing SDM, these results might reflect an initial savings when moving to a higher level of SDM, but a lack of a continued reduction in costs when high levels of SDM versus lower levels are maintained. With stable levels of SDM, disease severity may be the primary determinant of costs and the impact of SDM may be minimized.23 Prospective studies will be needed to confirm and explain these results in more detail.
In contrast to total health care expenditures, out-of-pocket costs reflect the financial burden for affected families. We found that out-of-pocket costs dropped by $142 over time for those with increasing versus decreasing SDM. In SDM, treatment decisions explicitly account for families’ values and preferences,1 including those related to out-of-pocket cost. In fact, some have advocated for SDM as a way to help clinicians acknowledge and address families’ financial concerns.24 Our findings suggest that when SDM increases as opposed to declines, families and clinicians may reach decisions that are less financially burdensome. Given that prior work has demonstrated that more than half of CSHCN have high out-of-pocket expenses,12,13 SDM may prove to be an important approach to help many families manage the costs of caring for medically complex children.
Little work has examined the impact of SDM on child health care costs and utilization. Decreases in cost were observed in a quality improvement initiative that used telephone-based case management to foster doctor-patient communication, patient self-management skills, and SDM for high-risk adults and children.25 In that study, preferences for more conservative treatments among adults participating in SDM were largely responsible for cost reductions. Prior research on health costs and utilization for CSHCN has primarily focused on the benefits of care coordination and the medical home.26–28 The introduction of care coordination through ambulatory subspecialty clinics for children with chronic conditions was found to reduce hospital costs,29 and expanded community-based care coordination also resulted in savings.30 In addition, findings from the 2005–2006 National Survey of CSHCN indicate that those with access to the medical home and adequate care coordination have approximately half the odds of having more than $500 in out-of-pocket costs31 and report fewer financial problems.32 Our findings extend these results by showing that including families in SDM may reduce both total and out-of- pocket costs.
In terms of health care utilization, we found that increasing SDM was associated with decreased rates of hospitalizations, ED visits, and office visits. Children with increasing SDM had 1.7 fewer office visits per child per year compared with those with decreasing SDM. Those with increasing SDM also had relatively lower rates of office visits compared with those with unchanged high or low SDM. Reducing office visits matters in this context because, at least for children with asthma, half of costs result from nonurgent outpatient visits.33 Our results may also be reassuring to outpatient pediatricians concerned about the time investment required to engage families of CSHCN in SDM.34 However, further study is needed to understand whether the decrease in office visits was offset by an increase in telephone care.
This study had several limitations. Although we implemented a latent class analysis to group children into distinct patterns of SDM based on their responses to 7 items corresponding to the definition of SDM,1 additional items might have allowed us to more fully characterize SDM. SDM exists between the extremes of paternalistic decision-making by the doctor alone and informed decision-making by the patient or family alone.1 Our study measure limited us to assessing only the extent of family involvement in decision-making, however, not who ultimately made decisions. In addition, we relied on household report as opposed to the direct observation of SDM. As a result, we could not verify how options were presented. Finally, although our data set provided a national perspective and detailed measures of cost and utilization, trials are needed to definitively assess how SDM affects costs and utilization in specific clinical contexts. Given the known high costs to pediatric practices of providing care coordination for CSHCN,35 studies should evaluate in more detail both the total reduction in expenditures as well as the costs to pediatric offices associated with implementing SDM.
We found that increasing SDM is associated with decreased health care costs and utilization for CSHCN. Results support prospective studies to determine if pediatric interventions to foster SDM reduce the financial burden of caring for these children.
This research was supported by an Academic Pediatric Association Young Investigator Award. In addition, the project described was supported by Award Number K23HD059919 from the Eunice Kennedy Shriver National Institute of Child Health & Human Development.
We thank Cyndi Ritz Wallin and Gary Moore of Social & Scientific Systems, Inc, and Dingwei Dai of The Children’s Hospital of Philadelphia Research Institute, Healthcare Analytics Unit, for their help with data preparation.
- Accepted September 7, 2011.
- Address correspondence to Alexander G. Fiks, MD, The Pediatric Generalist Research Group, The Children’s Hospital of Philadelphia, 3535 Market Suite, Room 1546, Philadelphia, PA 19104. E-mail:
The content is solely the responsibility of the authors and does not necessarily represent the official views of the Eunice Kennedy Shriver National Institute of Child Health & Human Development or the National Institutes of Health.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
Funded by the National Institutes of Health (NIH).
- Institute of Medicine (US)
- ↵The Patient Protection and Affordable Care Act. Public Law No: 111-148, 124 Stat 1025 (2010).
- McPherson M,
- Weissman G,
- Strickland BB,
- van Dyck PC,
- Blumberg SJ,
- Newacheck PW
- American Academy of Pediatrics Medical Home Initiatives for Children With Special Needs Project Advisory Committee
- Sia C,
- Tonniges TF,
- Osterhus E,
- Taba S
- McPherson M,
- Arango P,
- Fox H,
- et al
- Parish SL,
- Shattuck PT,
- Rose RA
- ↵Agency for Health Care Research and Quality. Medical Expenditures Panel Survey. Available at: www.meps.ahrq.gov/mepsweb/survey_comp/household.jsp. Accessed April 15, 2010
- ↵Agency of Health Care Research and Quality. Medical Expenditures Panel Survey: MEPS Topics: Health Care Costs/Expenditures. Available at: www.meps.ahrq.gov/mepsweb/data_stats/MEPS_topics.jsp?topicid=5Z-1. Accessed March 23, 2011
- Fiks AG,
- Localio AR,
- Alessandrini EA,
- Asch DA,
- Guevara JP
- Goodman L
- Landgraf J,
- Abaetz L
- ↵Machlin S, Yu W, Zodet M. Computing Standard Errors for MEPS Estimates. Agency for Healthcare Research and Quality. Rockville, MD; 2005. Available at: www.meps.ahrq.gov/survey_comp/standard_errors.jsp. Accessed March 30, 2011
- Ziring PR,
- Brazdziunas D,
- Cooley WC,
- et al.,
- American Academy of Pediatrics, Committee on Children with Disabilities
- American Academy of Pediatrics Ad Hoc Task Force on Definition of the Medical Home
- Homer CJ,
- Klatka K,
- Romm D,
- et al
- Turchi RM,
- Berhane Z,
- Bethell C,
- Pomponio A,
- Antonelli R,
- Minkovitz CS
- Lozano P,
- Fishman P,
- VonKorff M,
- Hecht J
- Fiks AG,
- Hughes CC,
- Gafen A,
- Guevara JP,
- Barg FK
- Antonelli RC,
- Antonelli DM
- Copyright © 2012 by the American Academy of Pediatrics