Appendix chapter 6
7.3 Selection of subgroups
This section analyzes an insurer's incentives to attract or deter specific sub-groups under the assumption that the regulator wants to retain an overall level of 80% for an insurer's incentives for efficiency". The easiest way to achieve this level of efficiency is to apply proportional risk sharing with a weight of 0.2 on actual costs. This reduces the predictable profits and losses for each sub-group by 20 %. As alternatives the regulator may apply risk sharing for high-risks for 3 % of the members, risk sharing for high-costs for 0.5 % of the members or outlier risk sharing with a threshold of Dfl. 40,000 (see Table 7.3).
35 The appendix shows additional results under the assumption that the regulator wants to retain 65 % of the overall incentives for efficiency.
166
7.3 Selection of subgroups
Table 7.6 Mean result in 1993 for subgroups formed on the basis of certain costs in 1991
N DEMO RSHR RSHC ORS PRS
(%) (3%) (0.5%) (40,000) (0.2)
Total cost 1991
0 14.3 883 662 661 768 706
1-2,164 75.7 266 83 132 191 213
2,165-4,651 5 -1,843 -1,524 -1,484 -1,708 -1,474
4,652-17,918 4 -2,978 -1,371 -2,098 -2,504 -2,382
> 17,918 1 -11,673 -2,684 -3,679 -6,899 -9,338 Prescribed d1'llgs 1991
0 28.4 711 516 514 614 569
1-619 61.5 231 77 118 166 185
620-1,098 5 -1,574 -1,471 -1,106 -1,405 -1,259
1,099-2,451 4 -3,633 -2,152 -2,635 -3,045 -2,906
>2,451 1 -12,118 -3,470 -5,894 -8,526 -9,694
Paramedical services 1991
0 83 212 130 152 178 170
1-336 7 -491 -407 -424 -482 -393
337-621 5 -694 -427 -418 -503 -555
622-1,532 4 -1,063 -820 -939 -1,033 -850
> 1,532 1 -6,403 -2,540 -3,791 -4,722 -5,122
N=47,21O. DEMO=demographic model. RSHR=demographic model + risk sharing for high-risks. RSHC=dcmographic model + risk sharing for high-costs. ORS=demographic model +
outlier risk sharing. All results arc statistically significantly different from zero, two-sided t-test. p<O.05.
Althongh the previons section showed that risk sharing for high-risks then yields the greatest reduction of the overall predictable profits and losses, the results may be different with respect to different subgroups. Table 7.6 shows the
7. Risk sharing as a sllppiemelll to demographic capitation
predictable profits and losses for the subgroups formed on the basis of certain costs in 1991. For instance, for those with the highest costs in 1991, the mean predictable loss is reduced to Ofl. 2,684 under risk sharing for high-risks, to Ofl. 3,679 under risk sharing for high-costs and to Ofl. 6,899 under outlier risk sharing. Under proportional risk sharing the predictable loss is Ofl. 9,338.
Similar results are found for those with the highest costs for prescribed drugs in 1991 and for those with the highest costs for paramedical services in 1991.
For those without any costs in 1991 or those without costs for prescribed drugs or paramedical services, risk sharing for high risks and risk sharing for high costs yield similar remaining profits. These profits are lower than those under outlier risk sharing or proportional risk sharing.
Table 7.7 shows the subgroups formed on the basis of prior hospitalization data.
For instance for those who were hospitalized in at least three of the four preced-ing years, the remainpreced-ing predictable loss under risk sharpreced-ing for high risks is about Ofl. 2,600. Under risk sharing for high costs the remaining loss is about Ofl. 5,100. Under outlier risk sharing it is about Ofl. 8,100 and under propor-tional risk sharing Ofl. 9,900.
Table 7.8 shows the consequences for subgroups formed on the basis of health survey data. For almost all subgroups with people that suffer from certain chronic conditions, risk sharing for high-risks as well as risk sharing for high costs lead to larger reductions of the predictable losses than either outlier risk sharing or proportional risk sharing. For instance, for those suffering from a serious heart disease, risk sharing for high risks reduces the mean predictable loss to about Ofl. 1,900 and risk sharing for high costs reduces it to about Ofl.
2,000. Under outlier risk sharing and under proportional risk sharing, the remaining loss is about Ofl. 3,400. For those suffering from cancer, risk sharing for high-risks reduces the mean predictable loss to about Ofl. 1,200 and risk sharing for high-costs to Ofl. 2,000. Under outlier risk sharing the remain-ing loss is about Ofl. 3,500 and under proportional risk sharremain-ing it is Ofl. 4,500.
For those without any chronic condition, risk sharing for high risks reduces the predictable profit to about Ofl. 320 and risk sharing for high costs reduces it to Ofl. 380. The profits after outlier risk sharing and proportional risk sharing are 168
7.3 Selection of subgroups
about Oft. 450 and Oft. 420 respectively.
Table 7.7 Mean result in 1993 for subgroups formed on the basis of hospital admissions and diagnostic cost groups in previous years
N DEMO RSHR RSHC ORS PRS
(%) (3%) (0.5%) (40,000) (0.2)
Dec
ill 19920 93.3 305 80 156 224 244
I 2.3 -1,395 -781 -866 -1,173 -1,116
2 1.8 -4,179 -1,472 -2,344 -3,160 -3,343
3 0.9 -10,694 -1,382 -4,425 -7,420 -8,555
4+5 0.2 -22,601 -1,734 -7,546 -12,391 -18,081
Unknown' 1.5 -2,828 -947 -1,989 -2,430 -2,262
Highest
DeC
in the period 1989-19921 8.3 -417 -474 -414 -436 -334
2 5.6 -1,584 -1,114 -1,133 -1,418 -1,267
3 3.1 -5,732 -1,694 -2,614 -4,060 -4,586
4+5 0.4 -12,159 -2,803 -5,090 -7,412 -9,727
Unknown' 2,8 -623 -465 -673 -680 -498
No. years with hospitalization ill the period 1989-1992
0 79.8 468 226 281 371 374
1 15.5 -812 -620 -640 -755 -650
2 3.7 -3,262 -1,558 -1,959 -2,575 -2,610
3 or 4 1 -12,398 -2,593 -5,088 -8,143 -9,918
N~47.21O. DEMO~demographic model. RSHR~demographic model + risk sharing for high-risks. RSHC~deIl1ographic model + risk sharing for high·costs. ORS~demographic model +
outlier risk sharing. All results arc statistically significantly different from zero, two-sided t-test, p<O.05.
a) These persons were hospitalized but the diagnosis is not available.
7. Risk sharing as a supplemelll to demographic capitation
Table 7.8 Mean result in 1993 for subgroups formed on the basis of health survey data
N DEMO RSHR RSHC ORS PRS
(%) (3%) (0.5%) (40,000) (0.2)
Presence of chronic conditions
None 61.1 523 322 375 451 418
At least one 38.9 -820 -504 -586 -708 -656
Asthma 5.0 -1,408 -1,187 -1,305 -1,428 -1,127
Heart disease 1.8 -4,330 -1,874 -2,034 -3,424 -3,464
Hypertension 6.9 -1,123 -322' -538 -775 -899
Diabetes 1.7 -2,895 -1,617 -1,752 -2,398 -2,316
Arthrosis 6.3 -755' -599 -770 -787 -604'
Rheumatism 2.9 -1,411' -833 -1,072 -1,176 -1,129'
Cancer 1.2 -5,602 -1,205' -1,954 -3,469 -4,482
Use of hOllle help 6r nursing
No 95.1 178 81 82 126 142
Yes 4.9 -3,487 -1,589 -1,608 -2,479 -2,790
Use of alternative practitioner
No 90.8 49' 55' 66 58' 39'
Yes 9.2 -475 -529 -637 -553 380
Education
Low 58.3 -50' -44' -42' -50' -40'
Medium 28.0 37' 31' 23' 32' 30'
High 10.3 249 138 118' 192 199
Unknown 3.4 -209' 91' 172' 10' -167'
N=IO,553. DEMO=demographic model. RSHR=demographic model + risk sharing for high-risks. RSHC=demographic model + risk sharing for high-costs. ORS=demographic model + outlier risk sharing.
') The mean result is not statistically significantly different from zero, two-sided t-test, p>O.05.
170
7.3 Selection of subgroups
Similar results are found for the subgroups formed on the basis of the use of home care. For the subgroups formed on the basis of the consultation of an alternative practitioner, the profits and losses are hardly changed by risk sharing.
For the subgroups formed on the basis of education, the profits and losses do not differ statistically significantly from zero in most cases.
Summarizing the results support the conclusion of the previous section that, given a certain overall level of incentives for efficiency, risk sharing for high risks and risk sharing for high costs yield greater reductions of incentives for selection than either outlier risk sharing or proportional risk sharing.