Abstract
Method
Configuration of characteristics in relation to outcome
Stability of prediction
Optimum time of release
Age at release
Discussion
Bibliography
Author: James A. INCIARDI , Dean V BABST
Pages: 33 to 39
Creation Date: 1971/01/01
With an increasing number of addicts being placed in aftercare programmes, once they have completed institutionalized treatment, it is becoming necessary for administrators to have some way of predicting which groups of patients will successfully respond to different programme approaches.
The strategy described here consists of developing classifications of narcotic users that can be a guide to differentiating the probable aftercare response. Initial classifications were based on the actual experience of 1843 cases released to the supervision of the California Department of Correction in 1965. Such classifications were then tested for their predictive utility by comparing group outcome rates with that of 1,380 cases released the following year.
A method for suggesting the optimum time for the release of different groups of addicts is also discussed.
A substantial number of the narcotics users under treatment in this country receive care in an institutionalized setting. Involuntary confinement is often by civil commitment, and some criminal offenders who go through the courts receive enforced hospitalization or institutionalization which is usually for a relatively indefinite period, this being followed by a conditional release to aftercare, or parole. With increased concern over narcotic addiction in large urban centres, each year facilities and services for drug users are expanded; and with the more widespread view that addiction is a chronic, relapsing disease rather than a simple manifestation of deviant tendencies, offenders with histories of narcotics use are more readily offered services in hospitals instead of being put into penal confinement, as in the past. Consequently, greater numbers are annually becoming eligible for community based placements in aftercare and clinically oriented parole programmes.
* This paper is based upon a larger study undertaken by the New York State Narcotic Addiction Control Commission (see Babst, Glaser and Inciardi, 1969).
Yet it is difficult to accurately and consistently predict the proportions which will fail or violate the conditions of their release. Without such predictive ability, the planning of institutional and aftercare rehabilitation programmes continues to be problematic, since such planning must be based on assumptions as to how well each programme will succeed with varying groups of drug users and addicts.
A suggested approach in the accomplishment of this task is the classification of institutionalized narcotic addicts in a manner which can effectively differentiate them by a demonstrably valid estimate of their rates of future adjustment upon release. The basic model employed in the development of such a narcotic user typology is similar to that developed for the construction of parole prediction scales or experience tables.
Essentially, parole prediction refers to "the estimate of probability of violation or non-violation of parole on the basis of experience tables, developed with regard to groups of offenders possessing similar characteristics" (Lejins, 1962). An "experience table" summarizes the post-institutional experience of a given release group. The table classifies individuals by their admission characteristics, and uses only those characteristics which are the most differentiating as to post-release response or adjustment in an aftercare setting. The experience table is similar to an acturial table which calculates or estimates insurance risks, but rather than indicating mortality or morbidity rates for each type of person, the experience table points to violation, recidivism or relapse rates. The hope, then, is to exemplify how the parole prediction method can be employed as an evaluation strategy for predicting the post-release adjustment of institutionalized narcotic addicts.
Numerous statistical techniques have been utilized in the construction of the desired predictive categories. A configuration of items was selected for this application since it has been proven effective and does not involve highly technical and abstract coefficients or weights (Babst, Gottfredson and Ballard, 1968).
The first predictive case classifications were developed from data for 1,843 male parolees with a history of narcotic addiction released to the supervision of the California Department of Correction in 1965. 1 Each case was followed up for one year to determine parole adjustment. Such adjustment was defined as unfavourable if within one year of release the parolee was returned to prison for violation of parole, arrested for a new offence and not restored to parole supervision, absconded, declared criminally insane, died during and as a result of the commission of a crime or from an overdose of narcotics, or returned to the Narcotic Treatment Control Unit. All other outcomes were defined as favourable. By cross-tabulating the intake characteristics by parole outcome and employing the Mean Cost Rating (MCR) technique as a statistical index of predictive selectivity, 2 those variables most related to parole outcome were distinguished. Table 1 represents the initial cross-tabulation of intake data by outcome.
It can be readily observed that a prior criminal record reflected an unfavourable prognosis in that 67 per cent of the first offenders had a favourable parole outcome, while only 33 per cent of the most repetitive offenders were successful. Furthermore, prisoners released on parole for the first time responded more readily to the parole setting, than those released again to the community subsequent to re-imprisonment for parole violation.
On the other hand, success at prior employment expressed a somewhat favourable significance. Those who were employed for a period of six months or more with one employer prior to institutionalization evidenced a greater degree of post-release success than those who were not as consistently employed. Similarly, non-property offences appeared to have a relationship to successful outcome.
Although longer periods of incarceration had apparent positive effects on parole outcome, this issue is discussed separately, not in terms of the over-all risk of release for different types of drug users, but rather with a view toward the optimum time for releasing a user in a particular category.
1 The authors are indebted to the California Department of Corrections for the data used in this study, especially to Dr. Lawrence A. Bennett, Chief of Research, and Mrs. Dorothy Jaman, Associate Social Research Analyst.
2 The Mean Cost Rating (MCR) measures the extent to which an item or combination of items have the ability to differentiate between categories with high and low success rates. MCR has a maximum value of 1.0 when all cases are in categories with either 100 per cent success or 100 per cent failure. It has a minimum value of zero when all categories into which cases are classified have the same percentage of success or failure (for a more detailed explanation as well as the technical derivation of the MCR, see Duncan, Ohlin, Reiss and Stanton, 1953; Duncan and Duncan, 1955).
Although the data presented in table 1 highlight relationships between certain intake characteristics and successful parole outcome, the findings have little value for predicting success or failure since many of those cases possessing a "positive" attribute may also possess one or more negative attributes. While 48 per cent of those with no prior commitments had a favourable parole outcome for example (a higher success rate than the group mean of 37 per cent), 52 per cent (or more than half) nevertheless had unfavourable outcomes. Similar circumstances surrounding outcome favourability existed among those offenders with prior commitment records involving either one prison term, or one to two jail sentences or juvenile incarcerations. This situation provides few guidelines for determining parole selection or release eligibility criteria, yet a more accurate and selective framework can be obtained by examining the configuration of characteristics most related to favourable post-release responses.
By selecting those variables from table 1 which are more highly selective (those with the highest MCR value) and cross-classifying them, groups possessing configurations of the most favourable and least favourable characteristics begin to emerge. In table 2, longest period of employment with one employer and prior commitment record data were combined and cross-tabulated with parole outcome. By this process, groups of cases can be observed which have greater differentiation as to parole success. It can be seen that although first offenders continued to have the same success rate as previously described, a highly unfavourable risk group appeared. While only 32 per cent of those who worked less than six months with one employer were successful, and 33 per cent of the cases with two or more prison or three or more jail or juvenile incarcerations were successful, only 28 per cent of the cases with both of those intake characteristics were successful. Furthermore, the predictive selectivity of this configuration increased in that the MCR value of .190 was greater than that of either variable when analysed independently. By introducing additional variables, even greater differentiation can emerge provided that a sufficient number of cases remain in each group to maintain statistical significance.
Before any predictive classification can be utilized as a tool for anticipating the success or failure of groups of cases possessing specific personal and social characteristics, it must be rigorously tested regarding its predictive ability. This was accomplished by developing configural or experience tables for the 1,380 male parolees with histories of narcotic addiction released to the supervision of the California Department of Correction in 1966. The identical follow-up and classification procedures used for the 1965 cohort were maintained for the new group.
In table 3, a three variable cross-classification of prior commitment record, longest period of employment with one employer, and number of prior paroles on current sentence was tabulated by parole outcome for both cohort groups. In comparing the outcome rates, considerable stability can be found in the majority of categories. This was especially marked for the two groups of parolees who had been employed six months or longer with one employer. Although less stability existed for the pairs of categories containing cases with fewer than six months with the one employer, the same relative ranking tended to persist.
Characteristics |
Number paroled |
Percent with favourable outcome |
MCR |
---|---|---|---|
Total cases
|
1,843
a
|
37
|
-
|
1.
Number of prior paroles on current sentence
|
|||
None
|
690 | 48 |
.175
|
One or more
|
1,153 | 31 | |
2.
Prior commitment record
|
|||
No prior record
|
52 | 67 |
.143
|
1 prison or 1-2 jail or juvenile
|
393 | 49 | |
2+ prison or 3+ jail or juvenile
|
1,398 | 33 | |
3.
Number of months in institution
|
|||
37 months or more
|
538 | 43 |
.090
|
25 to 36 months
|
460 | 36 | |
13 to 24 months
|
574 | 35 | |
12 months or less
|
269 | 32 | |
4.
Age at release
|
|||
40 years and over
|
342 | 41 |
.082
|
30 to 39 years
|
1,017 | 31 | |
Under 30 years
|
484 | 44 | |
5.
Longest period of employment with one employer
|
|||
6 months or more with one employer
|
1,175 | 40 |
.081
|
Less than 6 months with one employer
|
668 | 32 | |
6.
Type of offence
|
|||
Not cheques, forgery or burglary
|
1,265 | 40 |
.071
|
Cheques, forgery or burglary
|
576 | 32 | |
7.
Escape record
|
|||
No escape record
|
1,459 | 39 |
.055
|
Prior escape record
|
384 | 31 | |
8.
Alcohol involvement
|
|||
Yes
|
496 | 39 |
.021
|
No
|
1,341 | 37 | |
9.
Grade placement
|
|||
9.5 grade or higher
|
306 | 45 |
.014
|
9.4 grade or less
|
1,202 | 43 |
aNot all items add to 1,843 since cases on which information was unknown on a particular item were not included in the classification of that item.
Characteristics |
Number paroled |
Percent with favourable outcome |
|
---|---|---|---|
Total cases
|
1,843
|
37
|
|
No prior commitment
|
6+ months with one employer
|
39 | 72 |
Never 6 months with one employer
|
13 | 62 | |
One prison or 1 or 2 jail or juvenile
|
6+ months with one employer
|
266 | 50 |
Never 6 months with one employer
|
127 | 46 | |
Two or more prison or 3+ jail or juvenile
|
6+ months with one employer
|
870 | 36 |
Never 6 months with one employer
|
528 | 28 | |
MCR = .190
|
Number paroled |
Percent with favourable outcome |
||||||
---|---|---|---|---|---|---|---|
Characteristics |
1965 |
1966 |
1965 |
1966 |
Percentage point difference |
||
Total cases
|
1,843
a
|
1,380
|
37
|
42
|
5
|
||
No prior commitment
|
6+ months with one employer
|
39 | 28 | 72 | 69 |
-3
|
|
Never 6 months with one employer
|
13 | 8 | 62 | 78 | 16 | ||
One+ prison or 1-2 jail or juvenile
|
6+ months with one empl.
|
First parole
|
125 | 132 | 63 | 64 | 1 |
Reparole
|
141 | 102 | 42 | 40 |
-2
|
||
Never 6 months with one emp.
|
First parole
|
62 | 44 | 45 | 43 |
-2
|
|
Reparole
|
65 | 45 | 46 | 40 |
-6
|
||
Two+ prison or 3+ jail or juvenile
|
6+ months with one empl.
|
First parole
|
289 | 242 | 50 | 53 | 3 |
Reparole
|
579 | 462 | 29 | 33 | 4 | ||
Never 6 months with one empl.
|
First parole
|
181 | 97 | 35 | 42 | 7 | |
Reparole
|
347 | 218 | 25 | 35 |
10
b
|
||
1965 parolees: MCR = .270
|
1966 paroles: MCR = .218
|
a Totals to 1,841 since 2 cases with unknowns were excluded.
b Difference in outcome rates for designated comparison is statistically significant at the .05 level using chi-square test.
Since narcotic users who are sent to prison or put under treatment in hospitals, or other types of institutional facilities, will ultimately be released (except for a small fraction who die during confinement), the more important question for criminal or civil commitment policy is not which cases should be released, but rather, when should a given type of narcotic user be released. This question is crucial since the costs of maintaining a patient in an aftercare programme can be as little as 10 per cent of the costs of institutionalized treatment (Babst and Mannering, 1965). Paramount to any programme then, is the determination of which types of cases can be more successful on aftercare if released earlier.
Item 3 in table 1 suggested that the longer the narcotic parolees were confined, the more favourable was their outcome. It must be noted, however, that items 1 and 2 manifested even greater differentiation of cases by outcome rates. Furthermore, numerous studies of prison populations have showed that favourable outcome rates most often decline with increased time served (Glaser, 1964: 301-303; Babst and Magnuson, 1965; Garrity, 1961: 366-370; Washington State Department of Institutions, 1964).
By tabulating the cross-classifications of each variable with the number of months in the institution and by parole outcome, several relationships of time to outcome rates were isolated. In table 4, for example, which viewed the nexus of time and outcome with employment and commitment record, a pattern became apparent. For those with better employment records and fewest prior commitments, favourable outcome rates were lowest when confinement was over three years or less than one year, suggesting that the optimum confinement period ranged from a minimum of 13 to a maximum of 36 months. For those who were associated with one employer for the period of six months prior to confinement, and who had numerous prior commitments, three or more years of confinement appeared to be the optimum. In general, number of months in an institution, when compared with classifications, had little consistent relationship to parole outcome.
Predicting the post-release adjustment of institutionalized narcotic addicts 37
Longest period of employment |
Prior commitment record |
Number of months in institution |
Number of cases |
Percent with favourable outcome |
---|---|---|---|---|
Total cases
|
1,843
a
|
37
|
||
Longest job 6 months or more with one employer
|
None or 1 prison or 1-2 jail or juv. inst. term
|
37 or more
|
74 | 41 |
25-36
|
87 | 59 | ||
13-24
|
99 | 52 | ||
12 or less
|
45 | 31 | ||
2 or more prison or 3 or more jail or juv. inst. terms
|
37 or more
|
231 | 46 | |
25-36
|
200 | 34 | ||
13-24
|
286 | 31 | ||
12 or less
|
151 | 32 | ||
Longest job less than 6 months with one employer
|
None or or 1 prison or 1-2 jailjuv. inst. terms
|
37 or more
|
52 | 54 |
25-36
|
30 | 43 | ||
13-24
|
38 | 53 | ||
12 or less
|
20 | 45 | ||
2 or more prison or 3 or more jail or juv. inst. terms
|
37 or more
|
181 | 32 | |
25-36
|
143 | 24 | ||
13-24
|
151 | 26 | ||
12 or less
|
53 | 28 |
a Totals to 1,841 since 2 cases with unknowns were excluded.
There has been some speculation as to a "maturation process" for narcotic addicts. Maturing out of addiction refers to the process by which addicts cease drug use and drug-seeking behaviour, as the problems for which they originally began taking drugs become less salient and less urgent (Winick, 1962). This would suggest that older addicts might be less likely to relapse as their age increases, but a recent test of this maturation hypothesis found that a majority of the cases studied were still involved in drug using behaviour, and that the extent of social impairment and criminality had tended to increase with advancing age (Ball and Snarr, 1969).
Measuring maturation is difficult due to the many factors and contingencies involved, yet an attempt was made with the present data by comparing older and younger subjects within each of the derived classifications. As many classifications were used in table 5 as the number of cases and differentiation ability would permit. Cases in each class were divided at a mid-point, under 35 years and 35 years or older, and outcome rates were computed.
Although the largest percentage of older addicts were found in the poorer risk groups, this was expected since the older the individual the more likely he was to have a more extensive prior criminal record; and this latter factor, as indicated earlier, was related to unsuccessful parole outcome. The analysis indicates, however, when comparing older cases with younger cases, substantially different outcome rates nevertheless existed. In six of the seven comparison classifications for 1965 and five of the seven comparison classifications for 1966, the older addicts had a higher percentage with favourable outcomes.
Data in the parent study (Babst, Glaser and Inciardi, 1969) were used for a special tabulation for this study. In these special tabulations, a similar pattern was found among the 424 addict-parolees released to the supervision of the New York State Division of Parole during 1966. Using similar classifications and with a one year follow-up period, it was found that in five of the six classifications the older subjects manifested a higher percentage of favourable outcome.
By applying the parole prediction model, relatively accurate statistical indices can be developed which relate social history and case characteristics to treatment response. It should be noted however, that classifications of this type do not predict individual response. Rather, predictive classifications indicate the relative proportion of individuals possessing similar characteristics that will respond in a certain way. As such, prediction tables do not assert that a specific case will succeed or fail; they predict relative success for groups. Viewed in this light, they suggest which categories have prospects of success or failure markedly above and below the average, and which categories have little or no prediction significance (Glaser, 1962).
Number paroled |
Percent with favourable outcome |
||||||
---|---|---|---|---|---|---|---|
Prior record |
Employment |
Type of parole |
Age at release |
1965 |
1966 |
1965 |
1966 |
No prior commitments
|
-
|
-
|
35 years or more
|
9 | 1 | 78 | 100 |
Less than 35 years
|
43 | 37 | 67 | 70 | |||
1 prison or 1 or 2 jail or juvenile
|
More than 6 months with one employer
|
-
|
35 years or more
|
54 | 38 | 67 a | 63 |
Less than 35 years
|
212 | 196 | 46 | 52 | |||
Never employed 6 months consecutively
|
-
|
35 years or more
|
9 | 6 | 67 | 33 | |
Less than 35 years
|
118 | 83 | 40 | 40 | |||
2 or more prison or 3+ jail or juvenile
|
More than 6 months with one employer
|
First parole
|
35 years or more
|
130 | 96 | 56 a | 59 |
Less than 35 years
|
159 | 146 | 44 | 49 | |||
Reparoles
|
35 years or more
|
281 | 246 | 30 | 29 a | ||
Less than 35 years
|
298 | 216 | 29 | 38 | |||
Never employed 6 months consecutively
|
First parole
|
35 years or more
|
115 | 58 | 38 | 43 | |
Less than 35 years
|
66 | 39 | 29 | 41 | |||
Reparoles
|
35 years or more
|
158 | 97 | 24 | 37 | ||
Less than 35 years
|
189 | 121 | 29 | 33 |
a Differences in outcome rates for designated comparisons is statistically significant at the .05 level using chi-square test.
In extending this technique for the purpose of determining the suitability of patients for varying treatment modalities, its increased usefulness becomes more apparent. The procedure evolves into something considerably more complex, however, since classifications must be based on data more inclusive than simple intake characteristics. In most applications, two components must be examined. The first embodies what the patient brought to the rehabilitative setting, including his basic social and personal characteristics, his drug abuse and treatment history, data relative to the onset of his drug career, and criminal history if any. The second component would include the treatment dynamics to which the patient was exposed. Essentially, by substituting "treatment received" for "number of months in institution" (table 4), one could determine which type of treatment would be most effective for which type of patient.
This process would be relatively simple if treatment modalities could be classified into categories as methadone maintenance, therapeutic community, narcotic antagonist, etc. But it becomes more complex when one considers that within each of these modalities, each patient is exposed to different sets of circumstances, varying intensity of a lternative components of the treatment programme, as well as changing sets of attitudes, values and relationships with both clinicians and other patients.
The strategy suggested, then, requires a complete pre-institutional history of the patient, as well as a standardized and uniform reporting system which is able to measure the nature and extent of the patient's exposure to all phases of treatment and institutional processes.
Babst, Dean V., and John W. Mannering (1965), "Probation versus imprisonment for similar types of offenders-a comparison by subsequent violations", Journal of Research in Crime and Delinquency 2 (July): 60-71.
Babst, Dean V., and Toni Magnuson (1965), A Study of the Relationship between Time served and Later Parole Violation Experience, Madison, Wisconsin Department of Public Welfare.
Babst, Dean V., Don M. Gottfredson and Kelley B. Ballard, Jr. (1968), "Comparison of multiple regression and configural analysis techniques developing base expectancy tables", Journal of Research in Crime and Delinquency 5 (January): 72-80.
Babst, Dean V., Daniel Glaser, and James A. Inciardi (1969), "Predicting post-release adjustment of institutionalized narcotic addicts: An analysis of New York and California Parole Experience and California Civil Commitment Experience", New York, New York State Narcotic Addiction Control Commission. Mimeographed.
Ball, John C., and Richard W. Snarr (1969) "A test of the maturation hypothesis with respect to opiate addiction". Bulletin on Narcotics 21 (October-December): 9-13.
Duncan, Otis Dudley, Lloyd E. Ohlin, Albert J. Reiss, Jr., and Harold R. Stanton (1953), "Formal devices for making selection decisions ", American Journal of Sociology 58 (May): 573-584.
Duncan, Otis Dudley and Beverly Duncan (1955), " A methodo- logical analysis of segregation indexes ". American Sociological Review 20 (April): 210-217.
Gerrity, Donald L. (1961), " The prison as a rehabilitation agency ", pp. 366-370 in Donald R. Cressey (ed.), The Prison. New York, Holt, Rinehart and Winston.
Glaser, Daniel (1962), " Prediction tables as accounting devices for judges and parole boards ". Crime and Delinquency 8 (July): 239-258.
Glaser, Daniel (1964), The effectiveness of a Prison and Parole System, Indianapolis: Bobbs-Merrill.
Lejins, Peter P. (1962), " Parole prediction - an introductory statement ", Crime and Delinquency 8 (July): 209-214.
Washington State Department of Institutions (1964), " Washington State adult correctional institutions ", Research Review No. XIV.
Winick, Charles (1962), " Maturing out of narcotic addiction Bulletin on Narcotics 14 (January-March): 1-7.