Handicaps, disabilities, and dependency in French prisons
Transcription
Handicaps, disabilities, and dependency in French prisons
Handicaps, disabilities, and dependency in French prisons Lessons from a special survey* National Prison Museum, housed in the former remand prison (maison d’arrêt) of Fontainebleau SICOM/Justice Ministry In 1997, Maud Guillonneau and Annie Kensey, two demographers at the Prison Administration Directorate of the French Ministry of Justice, noted the shortage of information on inmates’ health status. They stressed the fact that “while the handicapped are obviously a category with specific needs, there are no quantitative data on them.” This finding led to the preparation of a survey on handicaps, disabilities, and dependency (Handicaps-IncapacitésDépendance: HID, an acronym used in the rest of this article) in prisons. The third HID survey The HID-Prisons survey was designed by the French Institute for Demographic Studies (INED). It was an extension of earlier surveys of the same kind conducted by INSEE among persons living in welfare and healthcare institutions (1998) and in private households (1999). Disabilities were defined to include those of physical or psychic origin but also of cultural origin, such as reading and writing impediments, and language problems. The main goal was to measure their prevalence in the prison environment—which, as some evidence suggested, could be very high. To begin with, the prison population has aged significantly. There are two reasons for this trend: (1) a change in legislation on sex crimes and offenses that has led to the jailing of older adults; (2) longer sentences. Moreover, inmates often come from lower-income groups [Cassan, Kensey, and Toulemon, 2000]. As the HID-Households survey had effectively demonstrated [Mormiche and HID project group, Virtual tour: http//www.justice.gouv.fr/musee/indexation 2000], the distribution of disabilities is strongly characterized by social inequality. The presence of persons with disabilities in prison naturally raises the issue of their rehabilitation, as well as of their living conditions. To collect information on these aspects, the survey includes questionnaire modules on family environment, education, employment, and income. There are studies specifically devoted to psychopathology in the prison environment and on other conditions and behaviors highly prevalent among inmates such as AIDS, hepatitis C, drug addiction, and alcohol consumption. The value of the new survey lies in its generalist Courrier des statistiques, English series no. 11, 2005 approach and the opportunities for comparison with earlier HID surveys in welfare/healthcare institutions and private households.1 A collective enterprise INED formed a project group to design the HID-Prisons survey comprising representatives of the Prison Administration Directorate of the Ministry of Justice (Direction de l’Administration Pénitentiaire: DAP), the Directorate General for Health at the Health Ministry (Direction Générale * Originally published in Courrier des statistiques, French ed., no. 107 (Sept. 2003), pp. 43-54. 23 Aline Désesquelles de la Santé: DGS), DREES,1 and several research agencies such as the National Institute for Health and Medical Research (Institut National de la Santé et de la Recherche Médicale: INSERM) and National Center for Scientific Research (Centre National de la Recherche Scientifique: CNRS). INSEE provided logistical support, most notably its interviewer network: the Institute had acquired a rich experience from the HID-Institutions and HID-Households surveys, as well as the 1999 Family Survey [Cassan, Héran, and Toulemon, 2000], whose field had been extended to the prison population. Funding came from INSEE, DAP, DREES, and—of course—INED. Preparing the survey protocol23 4 5 The initial plan was to submit the HID-Prisons questionnaire directly to a sample of about 700 inmates. But, given the uncertainty over the frequency of disability status in prison, there was a risk of covering an insufficient number of persons effectively suffering from disability. Should the administration of the HID questionnaire be preceded by a screening operation to identify persons with disabilities—along the 1. The Directorate for Research, Studies, Evaluation, and Statistics (Direction de la Recherche, des Études, de l’Évaluation et des Statistiques: DREES) was established in 1998 at the Ministry of Employment and Solidarity. It is classified as the “ministerial statistical office” (Service Statistique Ministériel: SSM) in charge of the health and social-affairs sectors. At the time of writing (2003), it reported to two government departments: the Ministry of Social Affairs, Labor, and Solidarity, and the Ministry of Health, Family Affairs, and Handicapped Persons. 2. In the 1999 population census, a specific questionnaire on “everyday life and health” (Vie Quotidienne et Santé: VQS) was distributed by a sample of enumerators to about 400,000 people. See Mormiche 2000. 3. The prison population is highly mobile, with a monthly turnover of about 15%. We thus soon realized that the two operations should not be conducted more than a month apart. 4. For the Family Survey in prison institutions, a large majority of interviewers consisted of persons with prison experience such as prison visitors, former lawyers, and researchers working on the prison population. 5. Administration of HID questionnaire to anyone supplying a positive response to at least one question in the VQS-Prisons survey. 24 Prison: an institution “almost” like any other To allow comparability of results obtained for the prison population and the general population, we took care to ensure that the HID-Prisons questionnaire was as similar as possible to those used in the earlier HID surveys. In fact, it hardly differs from the HID-Institutions questionnaire. Prison is “almost” an institution like any other—with the obvious difference that inmates are not free to come or go as they please. We therefore had to adapt the wording of questions on travel; other questions such as those on purchases and vacations were, of course, eliminated. lines of the procedure applied in the HID-Households survey?2 And, if so, how and when should the screening questionnaire be administered? It was decided to conduct an initial test, whose results would give an idea of the usefulness and feasibility of prior screening via a self-administered questionnaire called VQS-Prisons (in French: Vie Quotidienne et Santé, i.e. Everyday Life and Health). The test was conducted in June 1999 in two penal institutions: the detention center in Nantes and the remand prison (maison d’arrêt) in Osny. It yielded two main lessons. First, the procedure was unsuitable: the response rates were a very disappointing 26% for Nantes and 32% for Osny. Second, disabilities were far more prevalent among inmates that in the total population. As a result, we saw no way to avoid a face-to-face administration of the questionnaires. However, there were three possible options: • direct administration of HID questionnaire: a very high prevalence of disabilities would make the screening operation pointless; • face-to-face administration of a screening questionnaire followed three weeks later3 by the HID questionnaire; • face-to-face administration of a screening questionnaire and HID questionnaire in the same session. We decided to test the third option. By comparison with the second, it admittedly had a drawback: we needed to choose a screening criterion that would be immediately detectable by interviewers and would thus have to be very simple. On the other hand, and more decisively, the third option offered many practical advantages: shorter mobilization of prison staff; no “losses” between screening and HID questionnaire administration; lower cost. The operation was conducted in January 2000 in three penal institutions: the remand prisons in Rouen and Amiens and the detention center in Villenauxe-la-Grande. Beyond the issue of the survey protocol, the trial run was designed to test the HID-Prisons questionnaire and to determine whether INSEE interviewers could conduct a survey in a prison environment.4 The response rates were fairly satisfactory: 76% in Amiens, 75% in Villenauxe and 69% in Rouen. Once again, the results showed an high prevalence of disabilities in prisons: 51% in Amiens, 60% in Villenauxe, and 62% in Rouen (admittedly, the screening criterion was rather unselective5). The reader will have noted that the locations with the highest response rates are those with the lowest apparent prevalence of disabilities: this finding led us to examine whether a selection bias was at work. The comparison between responses provided to the two questionnaires in sequence—VQSPrisons and HID-Prisons—revealed some discrepancies, confirming the need to set up a large “control sample.” As with the HID-Households survey, the control sample served a dual purpose: (1) to compare the status of persons with disabilities against that of other persons; (2) to “recapture” false “negatives,” i.e., persons effectively suffering from disabilities despite their not having been identified as such at the screening stage. Handicaps, disabilities, and dependency in French prisons The test also showed that INSEE interviewers would be perfectly suited to the task and that the average completion time of 32 minutes6 for the HID-Prisons questionnaire was totally acceptable. Final protocol Despite the strong prevalence of disabilities in the prison environment, we decided to maintain the screening operation, whose marginal cost seemed relatively modest. The noscreening procedure would, in fact, have required administering many more “control questionnaires” to obtain the same final number of “positive” questionnaires. This would inevitably have obliged interviewers to perform more painstaking checks, not to mention the probable impact on response rates. In practice, all inmates in the survey sample were first given the VQSPrisons questionnaire. All those responding positively to one of the questions on presence of disabilities, activity impediments, handicaps, or need for aid linked to a health problem (questions numbered 3-17: see facsimile of questionnaire pp. 3133) were invited to respond to the HID questionnaire. To construct the control sample, one in three inmates were also invited to participate irrespective of their responses to the VQS questionnaire. As noted earlier, given the expected failure rate, there was a major risk that a selection effect could significantly bias the results. Inmates in good health might choose not to participate on the grounds that the survey did not concern them; conversely, inmates in poor overall health might be excluded because of their difficulty in going to the visiting room. To check this bias risk, we set up a procedure for collecting additional information. Doctors in participating institutions agreed to conduct an individual assessment of the general health status of all inmates initially selected, regardless of whether or not they later responded to the VQS questionnaire. The assessment used a four-level scale (excellent / good / poor / very poor) and was typically performed by examining medical records. The information gathered, collated with the separately available information on survey participation, would allow a test of the hypothesis of sample selection bias. Two-stage sample construction Institutions were sampled on the basis of the prison population statistics at April 1, 2000. At that date, there were 174 penal institutions in metropolitan France (mainland + Corsica), ranging in size from only 15 inmates to over 3,000. To spare the UCSA7 doctors an excessive work burden, it was agreed that no more than a hundred or so inmates should be surveyed in each institution visited. Conversely, given the desirability of assigning at least two interviewers to each institution, we would need to set a minimum number of interviews to be conducted in each institution. We eventually opted for a sampling of 50 or 100 inmates per institution. We therefore excluded the nine institutions with fewer than 50 inmates from the scope of the survey. By contrast, we specified the inclusion of the four institutions with over 1,000 inmates: three in the Paris area (Fresnes, La Santé, and Fleury-Mérogis) and one in Marseille (Les Baumettes). All four prisons host a population significantly different from the rest of the prison population in many ways: age distribution, structure by socio-economic status, proportion of aliens, type of offenses, etc. The other institutions, insofar as they were located in areas covered by the nine INSEE Regional Offices involved in conducting the survey,8 were distributed across six strata by cross-tabulating two variables: (1) number of inmates in the institution (three categories: small institutions with 50-149 inmates, medium-sized institutions with 150349 inmates, and large institutions with 350-999 inmates); (2) category of institution (two categories: Courrier des statistiques, English series no. 11, 2005 (a) maison d’arrêt [remand prison] (b) centre de détention [detention center] or maison centrale [highsecurity prison]). Next, we conducted an equal-probability sampling in each of the six strata thus defined. This yielded 28 institutions, which we added to the four very large institutions mentioned earlier, all of which are maisons d’arrêt. The 28 institutions sampled at random consisted of 21 maisons d’arrêt (7 small, 8 medium-sized, 6 large) and 7 centres de détention and maisons centrales (2 small, 1 medium-sized, 4 large).5678 The total number of inmates to be surveyed in these 32 institutions was set at 2,800, of whom 550 in small institutions, 850 in mediumsized institutions, and 1,400 in large/very large institutions. On the Friday preceding the collection week, the interviewers selected them at random from the list of inmates then present in the institution.9 Three categories were excluded from the selection: (1) minors, who could not be questioned without parental authorization; (2) inmates on day parole, who could not have been reached at the interviewers’ authorized visiting hours; (3) hospitalized inmates, as our purpose was to identify chronic disabilities, and most disabilities treated in hospital are temporary. This choice was also consistent with the one made for the generalpopulation survey, since the HIDInstitutions survey was restricted to long-term-care wards. 6. The survey had to be carried out entirely by means of paper questionnaires, except for CAPI (computer-assisted personal interviewing) arrangements. This is because it would have been impossible to allow PCs to be brought into prison institutions. 7. Primary-care units (in full: Unités de Consultations et de Soins Ambulatoires). 8. The following Regional Offices were involved: Alsace, Champagne-Ardenne, Haute-Normandie, Île-de-France [Paris Area], Limousin, Nord-Pas-de-Calais, Pays de la Loire, Poitou-Charentes, and Provence-AlpesCôte d’Azur. 9. Naturally, all persons selected at random were informed by letter that they had been chosen to take part in the survey. 25 Aline Désesquelles Categories of French prisons SICOM/Justice Ministry Persons serving sentences of up to one year and persons on remand pending final sentencing are placed in remand prisons (maisons d’arrêt). Detention centers (centres de détention) receive persons sentenced to more than one year in jail and with the best prospects of rehabilitation. The most difficult inmates are held in higher-security prisons (maisons centrales). These various incarceration arrangements may be applied in separate wards (quartiers) within the same penal institution. our control sample. In all, therefore, 1,314 persons were supposed to respond to the HID questionnaire. At this stage, we registered 30 failures, of which 13 refusals to respond and 17 cases of inability to respond: most of the latter consisted of persons with inadequate command of French. General health status: no significant difference between respondents and non-respondents... We naturally examined the breakdown of VQS questionnaire respondents and non-respondents by general health status as assessed by UCSA doctors and, conversely, the distribution of persons whose health status had been rated excellent, good, poor, or very poor according to whether they took part in the survey or not. This examination (see tables) revealed no significant difference between respondents and non-respondents. ...but wide variations between institutions La Santé prison in Paris, built in 1867 by the architect Vaudremer Collection results The collection period ran from Monday May 14 to Friday May 18, 2001. The work was carried out by 64 INSEE interviewers, in teams of two per institution visited.910 Of the 2,800 persons in the initial sample,10 2,031 (1,951 men and 80 women) actually responded to the VQS questionnaire. The participation rate averaged 72.5%, ranging from 47% to 96% in individual institutions. Participation was generally higher in small institutions, where it is easier to mobilize staff and inmates: the rate there was 82%, versus 70% in institutions with more than 149 inmates. The refusal rate was 20.9% and the rate of non-response for other motives11 was 6.6%. Of the 2,031 VQS respondents, 950 responded positively to at least one of questions 3-17 and were therefore invited to respond to the HID questionnaire. We also selected an additional 364 persons to form Chart 1 shows the survey participation rate for each institution on the X-axis, and the proportion of persons whose general health status was rated poor or very poor by UCSA doctors on the Y-axis. The absence of linkage between the two variables is clearly visible, as is the very wide variation between institutions (from 3% to 58%) of the proportion of inmates in poor or very poor health. A previous study on prison entrants [Mouquet, 1999] revealed sharp disparities between institutions. In Number of institutions selected and inmates surveyed by institution size Institutions visited 10. I.e., a sampling rate of just over 6%. At May 1, 2001, the prison population included in the survey field totaled almost 44,000. 11. In most cases because the person had just been released or transferred to another facility (61 cases) or had been excluded in advance because (s)he posed a security risk (46 cases). 26 Inmates surveyed 550 Small institutions 9 Medium-sized 9 850 10 1,000 4 400 32 2,800 Large Very large Total Handicaps, disabilities, and dependency in French prisons Breakdown of participants and non-participants by health status General health status Excellent Good Poor Very poor Undetermined Total No 30.2% 41.5% 17.7% 1.4% 9.3% 100.0% Yes 33.3% 44.4% 17.5% 2.2% 2.6% 100.0% Total 32.5% 43.6% 17.5% 2.0% 4.4% 100.0% Participation Breakdown of inmates in each health-status category by participation/non-participation General health status Excellent Good Poor Very poor Undetermined Total No 24.9% 25.5% 27.1% 18.9% 56.3% 27.5% Yes 75.1% 74.5% 72.9% 81.1% 43.7% 72.5% Total 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% the present case, real differences very likely explain only part of the observed dispersion. Some of the variation probably stems from the procedure for assessing inmates’ general health status, which is more or less optimistic depending on the doctor. The 32 penal institutions surveyed may be roughly divided into three groups: a “median” group comprising 11 institutions in which more than three-quarters of inmates were classified in the two central categories (”good” and “poor”), and two “outlier” groups where the assessment scale appears to have shifted toward “excellent” and “very poor” respectively. It is easy to understand the relativity of doctors’ assessments. In an institution where most inmates are in good health, the doctor’s criteria for regarding an individual as being in very good health are undoubtedly stricter. These factors lead us to the following conclusion: to test the hypothesis of a selection effect, we cannot simply compare the general health status of survey participants and nonparticipants in the aggregate. We need to make the comparison for each institution. vivid illustration. The X-axis shows the proportion of inmates “screened” by the VQS questionnaire; the Yaxis shows the proportion if all inmates initially selected had actually taken part in the survey. We estimated the latter value as follows. Let us assume that, in a given institution, the proportion of respondents screened by the VQS questionnaire lies between the proportions of inmates whose general health status is assessed by the doctor as being (1) very poor and (2) poor or very poor. We can then determine the coordinates of the first proportion relative to the two A closer look at the sample others (barycentric coordinates). Assuming these coordinates are valid for non-respondents as well, we deduce (1) an estimated proportion of “screened” nonrespondents and (2) the overall result including respondents and non-respondents. As we can see on the chart, the points obtained align almost perfectly on the first diagonal. In each of the 32 institutions surveyed, therefore, the inclusion of non-participants has virtually no impact on the estimated proportion of screened persons, so the result of the VQS screening is not biased by non-response. Chart 1 General health status: Poor/Very poor Participation The absence of a significant difference between the general health status of respondents and non-respondents, which we observed for the overall sample, is also confirmed for each individual institution. Chart 2 offers a fairly Courrier des statistiques, English series no. 11, 2005 27 Aline Désesquelles % VQS+ estimated Chart 2 % VQS+ (participants) Chart 3 Assessing the assessment For sake of completeness, we need to make sure that the assessment of inmates’ general health status by UCSA doctors is a good proxy for the presence of disabilities. The information letter sent to prison doctors did, of course, provide guidelines for this, but they may not have been easy to follow. Chart 3 cross-tabulates, for each institution, the percentage of respondents screened by the VQS questionnaire (Y-axis) and the percentage of inmates whose general health status was assessed by the doctor as poor or very poor (X-axis). The fine linear relationship we hoped to see is, alas, not perceptible. But the non-correlation between the two indicators observed is not surprising if we accept that doctors in different institutions produced their assessments differently. % VQS+ Specificity of prison surveys General health status: Poor/Very poor Ensuring anonymity of information gathered In each institution, a warden was put in charge of managing the list of persons selected to take part in the survey (see facsimile p. 30). The warden summoned the inmates to the lawyers’ visiting room—the locale where the interview would take place. If the inmate refused or was unable to participate, the warden would check the appropriate box in the right-hand columns on the list. Because of this arrangement, the interviewer was unaware of the respondent’s identity, unless, of course, the respondent had voluntarily introduced himself/herself. The interviewer did not retrieve the list until the end of the final interview, in order to make sure that the list and the collection process were complete. After these checks, the interviewer was supposed to detach the right-hand part of the list concerning survey participation and send it to INED. The left-hand part was given to the UCSA chief physician who, after performing his or her assessment, destroyed the nominal data and sent the rest of the form to INED. The information on overall health status and on survey participation was matched by means of the bar code preprinted on the form. 28 The crucial point, no doubt, is inmate participation. As we saw, it can vary quite widely with the choice of survey protocol and with the institution. While we are indeed dealing with a captive population, that does not mean it is docile. Moreover, it is a population with a very fast turnover (see footnote 3). Lastly, given the various activities in which inmates participate (workshops, sports, walks, etc.), “obtaining” an interview sometimes requires perseverance. Optimal cooperation by prison staff is thus a vital necessity. In this respect, the organization of the HID-Prisons survey—despite its broadly very satisfactory outcome— would no doubt have been even more successful if contacts with selected institutions had been made earlier and the stakes of the operation had been more fully articulated. Another important issue is, of course, security, which encompasses not only the interviewers’security but also organizational problems due Handicaps, disabilities, and dependency in French prisons to prison security constraints. In practice, everything went as well as could be, including in the highsecurity prison (maison centrale) that we were able to include in the survey sample thanks to the excellent cooperation established with the Prison Administration Directorate (DAP). Our regret—which is also a lesson for future prison surveys—is that we did not plan more interviews in that type of institution, or at least a sufficient number to allow specific post-survey analyses. Inmates of high-security prisons form a population with very distinctive characteristics, both in penal and socio-demographic terms. confirmed their strong motivation and wide-ranging talent.11 In conclusion, we would like to pay tribute to INSEE interviewers. They had already undertaken two highly challenging operations—the HIDInstitutions survey and the Survey of Users of Shelters and Soup Kitchens12—on “tough” subjects, in unconventional environments, and using novel collection methods. The HID-Prisons survey brilliantly Institut National d’Études Démographiques (INED) Aline Désesquelles 12. See Cécile Brousse, Bernadette de la Rochère, and Emmanuel Massé, “The INSEE survey of users of shelters and soup kitchens: an original methodology for studying the homeless,” Courrier des statistiques, English series, no. 9 (2003). References Cassan F., Héran F. and Toulemon L., 2000, “Study of family history: France’s 1999 Family Survey,” Courrier des statistiques, English ed., no. 6, pp. 7-19. Cassan F., Kensey A., and Toulemon L., 2000, “L’histoire familiale des détenus,” INSEE Première, no. 706, April 2000. Désesquelles A. and HID-Prisons Project Group, 2002, “Le handicap est plus fréquent en prisons qu’à l’extérieur,” INSEE Première, no. 854, June 2002. Désesquelles A., (2005), “Disability in French Prisons: How Does the Situation Differ from that of the General Population?” Population-E, 60(1-2), 2005. Guillonneau M. and Kensey A., 1997, “La santé en milieu carcéral – Éléments d’analyse démographique,” Revue française des affaires sociales, no. 1, Jan.-March 1997, pp. 41-60. Mormiche P., 2000, “The INSEE survey on handicaps, disabilities, and dependency: aims and organization,” Courrier des statistiques, English ed., no. 6, pp. 21-32. [originally published in French in 1998]. Mormiche P. and HID Project Group, 2000, “Le handicap se conjugue au pluriel,” INSEE Première, no. 742, October 2000. Mouquet M.-C., 1999, “La santé à l’entrée en prison en 1997: un cumul des facteurs de risque,” DREES-Études et résultats, no. 4, January 1999. Courrier des statistiques, English series no. 11, 2005 29 Aline Désesquelles 30 Handicaps, disabilities, and dependency in French prisons Courrier des statistiques, English series no. 11, 2005 31 Aline Désesquelles 32 Handicaps, disabilities, and dependency in French prisons Courrier des statistiques, English series no. 11, 2005 33 Aline Désesquelles 34