Evaluating Child Health Plus in Upstate New York: How Much Does Providing Health Insurance to Uninsured Children Increase Health Care Costs?
Background. In response to the increase in the number of American children without health insurance, new federal and state programs have been established to expand health insurance coverage for children. However, the presence of insurance reduces the price of care for families participating in these programs and stimulates the use of medical services, which leads to an increase in health care costs. In this article, we identified the additional expenditures associated with the provision of health insurance to previously uninsured children.
Methods. We estimated the expenditures on additional services using data from a study of children living in the Rochester, New York, area who were enrolled in the New York State Child Health Plus (CHPlus) program. CHPlus was designed specifically for low-income children without health insurance who were not eligible for Medicaid. The study sample consisted of 1910 children under the age of 6 who were initially enrolled in CHPlus between November 1, 1991 and August 1, 1993 and who had been enrolled for at least 9 continuous months. We used medical chart reviews to determine the level of primary care utilization, parent interviews for demographic information, as well as specialty care utilization, and we used claims data submitted to CHPlus for the year after enrollment to calculate health care expenditures. Using this information, we estimated a multivariate regression model to compute the average change in expenditures associated with a unit of utilization for a cross-section of service types while controlling for other factors that independently influenced total outpatient expenditures.
Results. Expenditures for outpatient services were closely related to primary care utilization—more utilization tended to increase expenditures. Age and the presence of a chronic condition both affected expenditures. Children with chronic conditions and infants tended to have more visits, but these visits were, on average, less expensive. Applying the average change in expenditures to the change in utilization that resulted from the presence of insurance, we estimated that the total increase in expenditures associated with CHPlus was $71.85 per child in the year after enrollment, or a 23% increase in expenditures. The cost increase was almost entirely associated with the provision of primary care. Almost three-quarters of the increase in outpatient expenditures was associated with increased acute and well-child care visits.
Conclusions. CHPlus was associated with a modest increase in expenditures, mostly from additional outpatient utilization. Because the additional primary care provided to young children often has substantial long-term benefits, the relatively modest expenditure increases associated with the provision of insurance may be viewed as an investment in the future.
A growing number of American children do not have health insurance; at any time during the year >11 million children (14%) are uninsured.1–5 Despite their lack of insurance, most uninsured children do receive some medical care, which is paid for from a variety of sources. Physicians and hospitals often provide charity care to uninsured patients. Other costs are borne by the families themselves, who pay out-of-pocket, and by other patients who are charged more to compensate for the costs of treating the uninsured. In response to the large number of children without insurance, President Clinton signed into law the State Children's Health Insurance Program (SCHIP), which will provide up to $40 billion through the year 2007 for states to expand health insurance coverage for children.6 With SCHIP, the government has assumed responsibility for most of the costs of caring for children who are eligible for the program.
One of the effects of expanding health insurance coverage is toincrease costs—at least in the short term—as the new beneficiaries tend to increase their use of health care.7Previous research has found that children also use more outpatient services when their out-of-pocket costs are lower.8Accordingly, a program for insuring previously uninsured children is likely to stimulate the use of medical services. However, little research has focused on the additional short-term costs of providing insurance to children previously uninsured.
The focus of this article is to measure these additional short-term expenditures. We estimated the expenditures on the additional services used by children living in the 6-county Rochester, New York, region who enrolled in the New York State's Child Health Plus (CHPlus) program. This program was one of the first programs designed specifically for low-income children without health insurance who were not eligible for Medicaid.
New York State implemented the CHPlus program in August 1991. Children <13 years old who were New York State residents having a gross family income below 222% of the federal poverty level and having no equivalent coverage were eligible for CHPlus. Children enrolled in CHPlus received a standard health insurance benefit package that offered coverage for the services of a comprehensive set of primary and specialty clinicians. During the study period, CHPlus only covered outpatient services. Since that time, the New York State legislature has expanded the program both by increasing the eligibility age to children up to 19 years (effective January 1, 1997) and by including coverage for inpatient services consistent with SCHIP guidelines.9
Initially CHPlus was financed primarily through the Statewide Bad Debt and Charity Care Pool. Funding began at $20 million annually and has grown steadily; the state contribution is expected to be >$256 million in the 1999 fiscal year. Additionally, New York State can expect to receive approximately $200 million more for CHPlus from the federal block grant program.9 Insurance premiums paid by some families have been a supplementary source of funds, paid on a sliding scale; the annual premiums are determined by family income. Premiums range from no payment for incomes 166% of the federal poverty level to $100 per family for those with incomes between 166% and 222% percent of the federal poverty level. Children from families whose income exceeded 222% of the federal poverty level could also participate in CHPlus, but they had to pay the full premium price. During our study period, only .4% of the children participating in the program were ineligible for the subsidy, and they were excluded from this study.
Study Design and Data Ascertainment
The data for this analysis were generated by a study of CHPlus in the area of Rochester, New York. The study setting was a 6-county region in upstate New York that includes the city of Rochester; the region has a total population of approximately 1 million people. The analysis of the additional expenditures attributable to CHPlus was based on the same sample of children that was used in the analysis of the CHPlus effect on utilization10; the study methodology is described in detail in an accompanying article.11
We attempted to contact 3203 families, representing 3754 children 0 to 6.99 years old (younger children) and 1769 children 7 years and over (older children). Younger children were eligible for the main study and older children were included if results of screening for asthma were positive.11 Contact was successful for 2606 (81% of 3203) families; representing 3136 (84% of 3754) younger children, and 1472 (83% of 1769) older children. Of those with whom contact was successful, 188 (6% of 3136) younger children and 100 (7% of 1472) older children were ineligible because parents reported that they terminated CHPlus enrollment after <9 months. Parents of 715 (23% of 3136) younger children and 313 (21% of 1472) older children refused to participate. Parents of 2232 (71% of 3136) younger children and 1058 (72% of 1472) older children agreed to participate in the study.
Parent interviews 12 months after enrollment in CHPlus were completed for 1828 younger children. Screening interviews for 1058 older children yielded 82 older children who had asthma. Children 2 to 7 were also screened for asthma, yielding 97 younger children who had asthma. Thus, the total number of cases, combining younger and older children, was 1910. Medical chart reviews were completed for 1802 (94%) of cases. Results in this article are for younger children only who had claims, medical chart and survey data. Results for children with asthma are discussed solely in an accompanying article.12
From the parent interviews, information was obtained about the children's demographic characteristics and the providers of their health care services for the year before and the year during enrollment in CHPlus. Medical charts were reviewed at all primary care sites (164 practices, all 12 emergency departments [ED], and 6 public health clinics) in the area to ensure that we had captured a comprehensive set of measures of health care utilization. Inpatient admissions were deduced from ED charts because most admissions occur via the ED.11 Parent interviews were used to measure specialty utilization because these data were incomplete in medical charts. Interviews also provided additional health and demographic information about the enrollee and the socioeconomic characteristics of the family.
Analysis of Costs
Payment totals, taken from the claims data submitted to CHPlus for the year during enrollment, were used to calculate expenditures for those services covered by CHPlus (these excluded inpatient expenditures, which were not covered by CHPlus at that time). Claims data were requested for all of the children participating in the study; when parents gave their consent to participate in the study, they included consent for access to all claims data. All claims data came from Blue Cross and Blue Shield of the Greater Rochester area, the local CHPlus insurer. These data included information that allowed us to identify the children being studied and to link their claims to the enrollment, medical chart review, and interview data. We also had information about the designated primary care physician (the actual provider of care) the date of encounter, diagnoses (categorized usingInternational Classification of Diseases, Ninth Revision, Clinical Modification codes), and procedures (categorized using Current Procedural Terminology-4 codes). The claims data also had information regarding the expenditures associated with the use of services covered by CHPlus. They included the amount paid by CHPlus to providers, adjustments and denials, copayments by the patient, and amounts covered by non-CHPlus plans if the child had concurrent insurance.
Total expenditures were calculated as the sum of payments to providers (net of adjustments), copayments by the patient, and payments by any other insurer. To account for fee increases during the study period, which included claims from 1992 through 1994, all reported claims were inflated to 1994 real dollars based on the price series provided by the US Department of Labor's Bureau of Labor Statistics for the upstate New York area.13
The objective of this analysis was to calculate the effect of CHPlus on health care expenditures. Because we were unable to collect expenditures data for the period before enrollment in CHPlus, an indirect method was used to impute these expenditures. Using data from the year during enrollment, for which expenditures data were available, we estimated a multivariate regression model for the relationship between expenditures and a variety of types of service. This regression model allowed us to calculate the average change in expenditures associated with additional units of utilization for each service type. We then estimated separately the change in utilization associated with enrollment in CHPlus (the CHPlus effect), as reported in an accompanying article.10 Combining the relationship between expenditures per unit of utilization and the CHPlus effect on utilization enabled us to determine the CHPlus effect on expenditures.
The datasets used to estimate the relationship between utilization and expenditures included expenditures information from the claims files, utilization data from the medical charts for the year during enrollment in CHPlus, and demographic and socioeconomic characteristics from the parent interview. The interview also provided us with the number of specialists seen. We used multivariate regression analysis to establish the relationship between expenditures and utilization for the year after enrollment, controlling for the other factors that influenced expenditures. Utilization measures for the following types of services were used in the analysis: acute care visits, chronic care visits, ED visits, follow-up visits, nurse only visits, other visits, well-child visits, hospital admissions, and the number of specialists seen.
We include the number of hospital admissions in this analysis even though inpatient services were not covered to test for the existence of spillovers where a hospital admission might affect outpatient expenditures. Because expenditures were skewed by a small number of children for whom expenditures were extremely high, we transformed the expenditures during the CHPlus period by taking its natural logarithm. The regression model had the logarithm of annual expenditures as the dependent variables. To identify factors other than utilization that accounted for the remaining variation in expenditures, we included several other variables in the regression analysis; race, education, and income were all statistically insignificant once utilization was controlled for. Only the child's age and the presence of a chronic condition had explanatory power beyond utilization. Therefore, the independent variables included in the final model were the complete set of utilization measures for the enrollment period, the child's age, and the presence of any chronic conditions. We used the coefficient estimates to predict the additional expenditures associated with an additional unit of utilization for each person in the sample after correcting for the bias involved in transforming from logarithms of expenditures to expenditures.14 To find the total additional expenditures, we applied this average additional expenditure per additional unit of utilization to the estimates of the CHPlus effect on utilization.
Table 1 shows the average number of ambulatory visits for the year after enrollment in CHPlus and the number of ED visits, hospital admissions, and specialists seen. The major components of total utilization per child for the year after enrollment in CHPlus were acute care visits per child (an average of 3.27 visits per year) and well-child care visits (an average of 1.41 visits per year). On average, children enrolled in CHPlus had 6.98 units of utilization (primary care visits plus the number of hospital admissions and specialists seen), at an average annual expenditure of $394.88 per year.
To determine how the utilization levels were related to expenditure levels, we initially examined the correlation between the 2 variables. As demonstrated by Table 2, the primary determinant of expenditures was total utilization, which had the highest correlation with expenditures (r = .55). With the exception of nursing-only visits and hospital admissions, all utilization measures were significantly correlated with expenditures. It is not surprising that hospital admissions were not statistically significant because inpatient services were not covered under CHPlus.
The relationship between average expenditures and factors other than utilization was complex (see Table 3). As age decreased, average expenditures and total utilization increased significantly. On average, a child <1 year old had more than twice the utilization and average expenditures compared with a child between the ages of 6 and 7. There was a weak relationship between expenditures (and utilization) and the parent's level of education. Average expenditures and utilization were highest for children whose parents had less than a high school education and lowest for children whose parents had some college education. Children who had at least 1 chronic condition had average expenditures twice as high as children who had no chronic conditions, while their total utilization was >2.5 times higher. Such differences in the relationship between expenditures and utilization suggested that we needed to use a multivariate analysis to analyze expenditures.
We performed a multivariate regression analysis on the data and used the derived coefficients to calculate the mean change in expenditures associated with an increase of 1 unit of utilization (Table 4), controlling for age and the presence of chronic conditions. All of the changes in expenditure levels were positive with the exception of chronic conditions and nursing visits. All primary care visits had additional expenditures that were in a narrow range. The negative sign for the cost associated with an additional chronic condition visit reflects the absence of a statistically significant relationship between expenditures and the number of chronic care visits. The decreased expenditures associated with nursing visits could reflect the fact that nursing-only visits were associated with a reduction in visits to higher cost physicians. Not surprisingly, visits to the ED and the use of specialty physicians had the largest effects on total expenditures. Hospital admissions had a very small spillover effect on expenditures. The increase in expenditures associated with an additional unit of other types of visits ranged from $47 to $64.
Finally, we multiplied the average change in expenditures by type of service to the change in utilization for each service type associated with the provision of CHPlus (CHPlus effect). In this manner, we estimated the total increase in expenditures attributable to CHPlus to be $71.85 per child in the year after enrollment.
We have found that the provision of CHPlus health insurance to children between the ages of 0 and 7 increased health care expenditures for ambulatory services by approximately $70/year per child. The additional expenditure was primarily attributable to increases in primary care utilization that are likely to be beneficial.
The method used to estimate these additional expenditures confirms that expenditures for outpatient services are closely related to primary care utilization. Age and the presence of a chronic condition were found to affect expenditures as well. Children who had chronic conditions and infants tended to have more visits, but these visits were less expensive, on average. The values we found for the average expenditure associated with an additional unit of utilization were generally in the expected range. All but 2 of these average additional expenditures were >0, implying that an additional unit of utilization increased expenditures. Further, the additional expenditures for all primary care physician visits were within a narrow range ($47–$64) for each visit type.
A basic assumption of this analysis is that changes in expenditures are determined by changes in the number of primary care visits. This assumption would appear to be especially questionable because even outpatient expenditures are heavily influenced by services other than primary care. In fact, one could construct examples where the estimated model would not predict expenditures well; for example, patients who have high expenditures but with few primary care visits. Fortunately, according to these data, average expenditures do tend to be related closely to primary care utilization. When we summed the products of average utilization and the corresponding additional cost per unit utilization (derived from our multivariate regression analysis), the total was remarkably close to actual average expenditures. This suggests that expenditures and utilization are, in fact, related linearly over the entire range of values.
A more basic limitation of the study arises from concerns regarding its generalizability. First, the study was conducted in a single location in a 6-county area surrounding Rochester, New York—an area that has several unusual features. On the one hand, a relatively high proportion of the population has insurance; on the other, the Rochester area has a high rate of child poverty. However, comparison of these study results to those from a statewide evaluation of CHPlus shows that while the details differ, the pattern and magnitude of the changes in service utilization after enrollment in CHPlus are similar to those found across New York State.15 It is likely, therefore, that the cost effects would be similar as well. Furthermore, the average incremental cost per patient of CHPlus insurance calculated in this article may not be greatly affected by the proportion of the population who had baseline health insurance.
Second, we restricted our analysis to expenditures for outpatient services. Expenditures for inpatient services were excluded primarily because, in this study, CHPlus did not have a statistically significant effect on the rate of admission. This finding is an important difference between the Rochester area and statewide studies. The statewide evaluation found that CHPlus had significantly reduced hospital admissions.15 Because each hospitalization is so expensive, such an effort would change our results dramatically. For example, a reduction of 1 percentage point in the probability of a hospital stay with an average cost of $3000 would lower net costs by $30, almost half the additional costs attributable to CHPlus. Thus, our results probably represent a conservative estimate of the incremental cost of CHPlus insurance—in other areas of New York State it may have been much lower.
A related limitation is that CHPlus did not cover hospitalizations at the time of the study. Although it is possible that inpatient coverage may have affected inpatient utilization, studies from the RAND Health Insurance Experiment did not find cost-sharing to influence pediatric hospitalizations.16
Third, our study focused on children <7 years of age. We have no basis for extrapolating these results to older children. This age group was selected because their utilization rates tend to be higher than those of children in the 7- to 13-year-old range,17 suggesting that additional costs for older children would likely be no higher than those observed for the younger population. The implications of insurance for the adolescent population is more uncertain because their health needs are quite different.18
The provision of health insurance to previously uninsured childrendoes increase utilization and, therefore, health care costs. This cost increase is associated almost entirely with the increased provision of primary care. Almost three-quarters (73%) of the increase in outpatient expenditures is associated with increased acute and well-child care visits, precisely the areas of unmet need of care for uninsured children.3 A negligible proportion of the cost increase was associated with increased ED or specialty use (<1% of the increase). Thus, the cost increase is incurred in the course of improving access to care, the prime motivation for the creation of CHPlus9 and an area that is believed to be cost-effective.19,,20
Finally, it is reassuring to see that our results, a 23% increase in costs associated with a child going from no insurance to full insurance, are substantially the same as those observed in the RAND Health Insurance Experiment.7 This correspondence suggests the approximate magnitude of cost increases to be expected with the provision of health insurance. Because children are relatively low-cost and the benefits of better health may improve their life outcome substantially, it is easy to see why there was a consensus to fund SCHIP.
Our analysis shows that all of the benefits of increased access to care conferred by CHPlus are associated with some relatively modest additional expenditures. The additional services provided involved an increase of 23% in the expenditures that would have occurred in the absence of CHPlus, a total of $71.85 per child per year. Because the additional primary care provided to young children tends to have substantial long-term benefits, the relatively modest expenditure increases associated with the provision of insurance may be viewed as an investment in the future.
- SCHIP =
- State Children's Health Insurance Program •
- CHPlus =
- Child Health Plus •
- ED =
- emergency department
- ↵US General Accounting Office. Health Insurance for Children. Private Insurance Coverage Continues to Deteriorate. Washington, DC: US General Accounting Office; 1996. Publication No. GAO/HEHS-96-129
- ↵Monheit AC, Cunningham PJ. Children Without Health Insurance. US Department of Health and Human Services; 1992. Publication No. 93-0025
- ↵State Children's Health Insurance Program. Public Law 105-33, Balanced Budget Act of 1997
- Valdez RB,
- Ware JE,
- Manning WG,
- et al.
- Holl JL,
- Dick AW,
- Shone LP,
- et al.
- Holl JL,
- Szilagyi PG,
- Rodewald LE,
- et al.
- Szilagyi PG,
- Shone LP,
- Holl JL,
- et al.
- Szilagyii PG,
- Holl JL,
- Rodewald LE,
- et al.
- ↵Bureau of Labor Statistics. Consumer Price Indices. Washington, DC: Bureau of Labor Statistics; 1995
- ↵Szilagyi PG, Zwanziger J, Rodewald LE, et al. Evaluation of the Child Health Plus in New York State. Final Report to the New York State Department of Health. Albany, NY: Department of Pediatrics, University of Rochester; 1996
- Leibowitz A,
- Manning WG,
- Keeler EB,
- Duan N,
- Lohn KN,
- Newhouse JP
- ↵Haggerty RJ, Roghmann KJ, Pless IB. Child Health and the Community. New Brunswick, NJ: Transaction Publishers; 1993
- ↵Starfield B. The Effectiveness of Medical Care: Validating Clinical Wisdom. Baltimore, MD: Johns Hopkins University Press; 1985
- ↵Newacheck PW, Stoddard JJ, Hughes DC, Pearl M. Children's access to health care: the role of social and economic factors. In: Stein REK, ed. Health Care for Children. New York, NY: United Hospital Fund; 1997
- Copyright © 2000 American Academy of Pediatrics