Chapter 5—Assumptions in Substance Abuse Actuarial Studies

Actuarial data are rarely perfect. Often, data are missing, incomplete, or believed to be inaccu-rate. Some subgroups of the covered population may not be adequately represented in the data, or some services in the benefit package may be unique. In such cases, the actuary needs to make assumptions. The Washington State study provides several good examples.

Currently Uninsured Families

Under the Washington health care reform plan, coverage is to be extended to currently uninsured low-income families. Little is known regarding utilization by this group, since there is no insurance company to keep track of billings and these families are not eligible for medicaid. Their lack of current health insurance coverage probably discourages them from seeking care, so any data from other sources are suspect. In this case, the State's actuary initially proposed assuming that utilization will be 15 percent higher than the rate for currently insured families, because uninsured families are younger (hence closer to the mean age of chemical dependency patients), poorer (hence more likely to be chemically dependent), and more likely to post pone seeking admission because of cost (hence creating pent-up demand) than currently insured populations are. Washington State officials were very concerned that the assumed 15-percent increase might be too low. If the State adopted a plan based on such an assumption and utilization was dramatically higher than the forecast, the legislature might terminate chemical dependency coverage to control costs. The State and its actuary needed as accurate an estimate as possible for utilization by currently uninsured persons.

Also of concern was the possibility that currently uninsured persons may have been postponing needed treatment, but would suddenly show up once coverage is extended. This pent-up demand, or "woodwork effect," could raise treatment admissions dramatically, particularly over the short term.

Washington had received a contract from the Center for Substance Abuse Treatment to conduct a prevalence study statewide. Part of that study was a household survey of adults, which included a chemical dependency scale. The State also asked respondents whether they were currently insured and whether they had entered chemical dependency treatment in the past year. With these data, the State could estimate prevalence of chemical dependency among currently uninsured families and could compare that figure with estimated prevalence among insured families. Although prevalence (the rate of chemical dependency) is different from utilization (the rate of admission to treatment), it is reasonable to assume that the number of persons who enter treatment from any group (utilization) is a constant proportion of the number of persons who need treatment (prevalence). This implies that the ratio of utilization by uninsureds to utilization by insureds would be about the same as the ratio of prevalence among uninsureds to prevalence among insureds. Washington could replace the 15-percent assumption about increased utilization with the following formula:

The State hoped to use the prior treatment data to refine its assumptions about pent-up demand among the currently uninsured group. Pent-up demand is created by lack of access, and the household survey's measurement of different rates of access to treatment over the past year by insured and uninsured groups could help the State infer the rate at which persons in the uninsured group might have wanted treatment but have been unable to access it. Unfortunately, the number of completed surveys was too small at the time to measure differences in prior-year utilization between chemically dependent insureds and chemically dependent uninsureds. The State and the actuary still needed to guess at the rate of pent-up demand. The final loading factor chosen for pent-up demand was 25 percent for each of the first 2 years, with no load thereafter. Since the prevalence survey had already included pent-up demand as "current need," the 25-percent loading factor was converted into a reduction in demand after the exhaustion of pent-up demand (after 2 years). The computed equivalent reduction is 20 percent.

Washington's final utilization estimates for the currently uninsured group were 6.7 admissions per 1,000 uninsureds per year for the first 2 years and 5.4 per 1,000 thereafter (Table 5–A). The estimate for the first 2 years was 53 percent higher than the actuary's initial proposal of 115 percent of the utilization rate for insureds.

Medicaid Enrollees

Medicaid posed different problems. The Washington Medicaid Management Information System has demographic data on each medicaid enrollee, whether he or she seeks chemical dependency treatment or not. Medicaid data bases are very similar in design and function to insurance company data bases, so reliability, completeness, and validity are comparable. Utilization rates for medicaid enrollees can be reliably calculated for any services included in (and billed separately in) the medicaid benefit package. In short, medicaid offers what seems to be an ideal data base.

The drawbacks came from restrictions on medicaid benefits. Medicaid pays for hospital-based care and for outpatient treatment, but not for nonhospital residential care. This posed problems if residential care was to be part of the benefit package under health care reform.

In Washington, hospital-based substance abuse treatment is available only to pregnant chemically dependent women. Under managed care, nonpregnant persons could be admitted to hospital-based substance abuse treatment services, but it seemed reasonable to assume that cost containment concerns would hold overall hospital utilization to about the same level.

Medicaid data for outpatient treatment utilization could not be accepted with the same confidence. Due to funding limitations, the State instituted policies restricting access to medicaid-funded outpatient treatment. Although the level of funding changed nearly 2 years before the actuarial study was initiated, lags in implementation and in billing for medicaid services resulted in incomplete data reflecting the cost controls being available for actuarial analysis.

Use of the medicaid data base for the actuarial study was precluded when the State's medicaid program managers reported that it would take several months to generate the necessary reports. Such a delay would mean that the Health Services Commission would have to make its decision without the actuarial study. The combination of questionable data and delayed availability led the State to look for other means of generating net-cost-per-person-per-month estimates for the medicaid population. The State decided to use the method outlined above for the uninsured population to generate estimates for the medicaid population.

Medicare Patients

Medicare posed still different problems. Prevalence of chemical dependency in the medicare population is low, and Washington State had completed too few surveys in the household study to get a reliable estimate. Medicare utilization data are available in insurance data bases, but reported admission rates are extremely low. The State believed the low reported utilization was due to lower need for chemical dependency treatment compared with younger adults, to poor diagnoses by practitioners, and to a desire to shield elderly patients and their families from a chemical dependency diagnosis. Prevalence studies from other States provided an estimated prevalence of chemical dependency in the elderly population of between 25 and 60 percent of the rate among the general adult population.¹ Taking the upper prevalence estimate and assuming that the ratio of utilization to need is the same as that for insured adults generally, Washington would expect a utilization rate of 1.7 per 1,000, or about 45 percent of the general utilization rate. Expecting that some of the misdiagnosis and reluctance to refer would still remain despite health care reform, the State and the actuary reduced the assumed utilization for medicare-aged population to 1.5 per 1,000, equal to 40 percent of the rate for the entire insured population.

Note

¹ The Epidemiologic Catchment Area study (see Norman S. Miller, Beth M. Belkin, and Mark S. Gold, "Alcohol and Drug Dependence Among the Elderly: Epidemiology, Diagnosis and Treatment," Comprehensive Psychiatry 32, No. 2 (1991): 153–65). Although the study did not include a Washington site, a similar study in Spokane, found a prevalence rate equal to 60 percent of the general population rate (see Martin J. Jinks and Raymond Raschko, "A Profile of Alcohol and Prescription Drug Abuse in a High Risk Community Based Elderly Population," DICP, The Annals of Pharmacotherapy 24, No. 10 (1990): 971–75).

Back to Top