تنظیم ریسک در بیمه سلامت و اثر بلند مدت آن
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|25307||2010||10 صفحه PDF||سفارش دهید||8950 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Health Economics, Volume 29, Issue 4, July 2010, Pages 489-498
This paper seeks to create new insights when judging the impact different risk adjustment schemes may have on the incentive to select risks. It distinguishes risk types with high and low profit potential and estimates long-run profits associated with risk selection in four scenarios (no risk adjustment, demographic only, including prior hospitalization, and including prior hospitalization and Pharmaceutical Cost Groups). The database covers 180,000 Swiss individuals over 8 years, 3 of which are used for model building and 5, to estimate insurers’ profits due to risk selection in the four scenarios. While these profits prove to be very high without risk adjustment and still substantial with demographic risk adjustment, they become surprisingly low when the crude morbidity indicator ‘prior hospitalization’ is included in the formula. These results clearly indicate the need for health status-related risk adjustment in insurance markets with community rating, taking into account insurers’ planning horizon.
Enthoven's proposal for regulated competition between social health insurers (Enthoven, 1978) has been used as a blueprint for reform in several countries (see van de Ven et al., 2007). One example is Switzerland with its comprehensive mandatory coverage for all citizens, offered by some 80 competing nonprofit health insurers. The new law of 1994 calls for semiannual open enrollment and community-rated premiums within the same fund. Premium reductions for adults within a given fund are only possible for contractual differences such as a higher deductible. However, with every adult paying the same premium – within a given fund for the same type of contract – but expected health care expenditure (HCE) varying widely, strong incentives for risk selection are created in the absence of an adequate risk adjustment scheme. Although risk selection is illegal, its prevalence in Swiss social health insurance has been reported repeatedly (Beck and Zweifel, 1998 and Beck et al., 2003). As van de Ven and Ellis (2000, Section 2.5) argue, risk selection produces no benefits to society (unless a dynamic view is adopted, where the threat of being classified as an unfavorable risk in the future helps to reduce moral hazard). The objective of risk adjustment is to mitigate incentives for risk selection. To this end, insurers with a below-average share of female and elderly consumers have to contribute to the risk adjustment fund, while insurers with an above-average share receive a payment from the fund. So far, the different schemes have been judged mainly in terms of their ability to predict individual HCE 1 year ahead (Newhouse et al., 1989, van Barneveld et al., 2000 and Holly et al., 2003). This criterion is subject to at least two criticisms. First, risk selection is not costless to insurers. As pointed out by van Barneveld et al. (2000) as well as Zweifel and Breuer (2006), this means that they will invest in this activity only if expected profits exceed the cost. However, the regression criterion of minimizing squared prediction error with regard to HCE fails to take cost considerations into account. In our model, we address the problem of costly risk selection activities by restricting attention to selection profits and losses exceeding a given annual threshold. Second, Zweifel and Breuer argue that insurers who want to stay in business must have an eye on present values rather than one-period profits. This paper, then, follows the lead of Shen and Ellis (2002) by estimating expected profits attainable from risk selection, given the information available to the insurer. It therefore only models the classification of risk types, neglecting the problem of how to attract or deter types. However, if profits are large enough, strategies to perform risk selection will most certainly be developed by crafty insurers. However, unlike Shen and Ellis (2002), the present analysis assesses the impact of risk adjustment if insurers’ planning horizon exceeds 1 year. In an attempt to reflect longer planning horizons (which agree with insurers’ preference for long-run contracts and guaranteed renewability unless prevented by regulation (Pauly et al., 1998)), the period of observation for expected profits is extended to 5 years in the body of the paper. This permits to take into account the fact that a currently favorable risk may develop into an unfavorable one, switch to a competitor, or die. Conversely, an unfavorable risk may recover to become a favorable one in the future. In the theory of statistics, these effects are known as “regression towards the mean” (Welch, 1985 and Beck, 2004). The empirical relevance of the regression towards the mean effect is assessed by varying the planning horizon from one to 5 years. This is possible thanks to a panel data set covering some 180,000 individuals over 8 years. The reminder of this article is structured as follows. In Section 2, a model of the insurer's decision to select risks is formulated, which subsequently permits to calculate the financial reward from this activity. After a description of the risk adjustment schemes and the database in Section 3, the details of the empirical estimation are explained in Section 4. Results are presented in Section 5. They indicate that even a crude adjustment in the risk adjustment formula to take future HCE into account serves to neutralize incentives for risk selection to a substantial extent also over a longer planning horizon. The final Section 6 is devoted to a summary and conclusions.
نتیجه گیری انگلیسی
There is a broad consensus that given managed competition with community-rated premiums, risk adjustment becomes a necessary regulation of health insurance markets. However, while a purely demographic risk adjustment formula has been recognized as being insufficient, its precise specification has remained controversial. Most of the empirical literature describes and tests for the relationship between a set of morbidity indicators and HCE of 1 year. This short time horizon is in accordance with the fact that managed competition usually allows for annual open enrollment. However, insurers have a strong interest in long-term customers in view of considerable acquisition cost. Moreover, consumers likely would think twice before signing up with an insurer whose planning horizon is as short as 1 year. Given a longer term perspective, insurers’ strategies are influenced by two empirical facts, viz. decreasing precision of forecasts and regression to the (conditional) mean. The first fact causes the risk of misclassification to increase with a longer planning horizon. However, to the extent that the regression is to the conditional rather than the grand mean of HCE, the second fact may serve to increase the payoffs to risk selection when the planning horizon is extended. In this research, the effectiveness of risk adjustment schemes is assessed in the light of these considerations. Purely demographic risk adjustment already increases the likelihood of misclassification but does not sufficiently neutralize the longer term, systematic differences in HCE to decisively mitigate the incentives for risk selection. For this, it takes a more refined formula that includes prior hospitalization and possibly Pharmaceutical Cost Groups (PCGs) as risk adjusters. Gains from risk selection are estimated based on predicted HCE net of copayments, premiums, and risk adjustment payments, discounted to present value and weighted by the probabilities of death and of switching to a competitor to obtain expected values. These calculations are performed for four different risk adjustment schemes, viz. (0) no risk adjustment, (1) demographic risk adjustment, (2) demographic risk adjustment, with prior hospitalization added as a simple morbidity indicator, and (3) PCGs complementing scheme (2). For a sample of some 180,000 Swiss individuals, expected net present values conditioned on the risk adjustment scheme were calculated. However, these values must exceed the variable cost associated with risk selection effort in order to trigger action on the part of the insurer. Since this cost is unknown, an arbitrary but not unrealistic value of ±CHF 1000 p.a. ($ 800 as of 2007) serves as a threshold. Thus, the insurer is assumed to be indifferent with regard to risks whose contributions to expected profit fall within this interval. The risk adjustment schemes distinguished modify incentives for risk selection according to expectations. The better they reflect morbidity, the smaller the share of the insured population that constitutes favorable and unfavorable risks, respectively. Adjustment using only age and sex (type 1) causes the share of favorable risks to drop from 56 to 40 percent, the share of unfavorable ones, from 21 to 18 percent. With schemes of type (2) and (3), the figures for the favorable risks drop to 26 and 20, and for the unfavorable risks, to 17 and 18 percent, respectively. It also should be noted that the characteristics of the subgroups change dramatically with type of risk adjustment. Average age of favorable customers increases from 46 (type 0) to 71 years (type 3), while the share of those belonging to one or more PCGs (an indicator of chronic illness) increases from 3 to 47 percent. The success of risk selection efforts is reflected by the insurer's ability to lower premiums and hence gain market share. In the absence of risk adjustment, deterring all unprofitable risks is estimated to result in a 46 percent premium reduction over 5 years. This longer term competitive advantage is reduced to 16 percent under type (3) risk adjustment, which takes into account both prior hospitalization and membership in a PCG. Interestingly, this figure is in the same range as the premium reductions that may be offered for participation in a managed care alternative, which constitutes a product innovation. Thus, it may be argued that type (3) risk adjustment is effective enough to redirect insurers’ efforts from risk selection to product innovation. In addition, the risk of misclassification is a mere 7 percent for an insurer “chasing the good risks” as long as there is no risk adjustment but increases to 25 percent under scheme (3). Refined risk adjustment therefore becomes even more effective than indicated by the expected contribution to profit because it exposes insurers to increased uncertainty. In conclusion, this research suggests that risk adjustment can be designed in a way as to be effective enough in the longer term to redirect insurers’ efforts away from risk selection in favor of product innovation while using easily available information such as prior hospitalization and membership in pharmaceutical cost groups.