Economic Impacts of Supported Employment for Persons With Severe Mental Illness. Eric A Lati mer, PhD 1

Transcription

1 IN REVIEW Economic Impacts of Supported Employment for Persons With Severe Mental Illness Eric A Lati mer, PhD 1 Back ground: Most per sons with se vere mental ill ness pre fer com pet i tive to shel tered vo ca tional set tings. Sup ported em ploy ment (SE) has be come a clearly de fined model for help ing peo ple with se vere mental ill ness to find and main - tain com pet i tive jobs. It in volves in di vid u al ized and rapid place ment, ongoing sup port and as sess ment, and in te gra tion of vo ca tional and men tal health staff within a sin gle clin i cal team. Pre vi ous studies show that SE secures competitive em ploy ment much more ef fec tively than do other ap proaches. This re view fo cuses on its eco nomic im pacts. Methods: Studies re port ing some ser vice use or mon e tary out comes of add ing SE pro grams were identi fied. These out - comes were tab u lated and are dis cussed in nar ra tive form. Re sults: Five nonrandomized and 3 ran dom ized stud ies com pare SE pro grams with day treatment or tran si tional em - ploy ment pro grams. The in tro duc tion of SE ser vices can re sult in any thing from an in crease to a de crease in vo ca tional ser vice costs, de pend ing on the ex tent to which they sub sti tute for pre vi ous vo ca tional or day treat ment ser vices. Over - all ser vice costs tend to be lower, but dif fer ences are not sig nif i cant. Earn ings in crease only slightly on av er age. Con clu sions: Con verting day treat ment or other less ef fec tive vo ca tional pro grams into SE pro grams can be cost-sav - ing or cost-neu tral from the hos pi tal, com mu nity cen tre, and gov ern ment points of view. In vest ments of new money into SE pro grams are un likely to be ma te ri ally off set by re duc tions in other health care costs, by re duc tions in gov ern - ment ben e fit pay ments, or by in creased tax rev e nues. Such in vest ments must be mo ti vated by the value of in creas ing the com mu nity in te gra tion of per sons with se vere men tal ill ness. (Can J Psy chi a try 2001;46: ) Key Words: vocational rehabilitation, supported employment, individual placement and support, mental illness, psychiatric disability, costs, cost-effectiveness, economics In creas ingly so phis ti cated ef forts are being made to sup port the com mu nity in te gra tion of per sons with se vere mental ill ness. Work is an es sen tial el e ment of that in te gra tion. Sur - veys in di cate that most in di vid u als with se vere mental ill ness want to work in reg u lar em ploy ment set tings, earning the min i mum wage or better (1). More over, stud ies show that work in creases their self-es teem (2,3) and qual ity of life (4). Yet US data in di cate that 10 years ago, at least, more than three-quar ters of per sons with se vere men tal ill ness re mained un em ployed (5). Tra di tional ap proaches to pro vid ing this Manuscript received and accepted June Research Scientist, Douglas Hospital Research Centre, Verdun, Quebec; Assistant Professor, Department of Psychiatry, McGill University, Mont - real, Quebec; Associate Member, Department of Economics, McGill Uni - versity, Montreal, Quebec. Address for correspondence: Dr E Latimer, Douglas Hospital Research Centre, 6875 LaSalle Boulevard, Verdun, QC H4H 1R3 lateri@douglas.mcgill.ca pop u la tion with mean ing ful work have in volved shel tered or tran si tional em ploy ment, which rarely lead to com pet i tive em ploy ment (6,7). Over the past 15 years, a rad i cally dif fer ent ap proach called sup ported em ploy ment (SE) has emerged (8,9). Ini tially ap - plied to vo ca tional re ha bil i ta tion of per sons with phys i cal and de vel op men tal dis abil i ties (8), SE in volves plac ing cli ents di - rectly into com pet i tive jobs and then of fer ing them ongoing sup ports place-train, rather than train-place. During the 1990s, US re search ers for mal ized the SE model for peo - ple with psy chi at ric dis abil i ties, un der the la bel In di vid ual Place ment and Sup port (IPS) (10,11). This for mu la tion of SE bor rows from the As ser tive Com mu nity Treat ment (ACT) model (12) the prin ci ple that treatment and re ha bil i ta tion ser - vices should be in te grated within a sin gle clin i cal team. The term sup ported em ploy ment, ap plied to per sons with se - vere mental ill ness, is in creas ingly be ing con strued in a man - ner con sis tent with the prin ci ples of IPS (13). That is the sense in which it will be used through out this re view. Can J Psychiatry, Vol 46, August

2 August 2001 Economic Impacts of Supported Employment for Persons With Severe Men tal Illness 497 SE prin ci ples have been de scribed ex ten sively else where (14) and will only be listed here: the search for com pet i tive rather than shel tered em ploy ment; no, or min i mal, prevocational train ing; in te gra tion of vo ca tional and men tal health treat - ment within the same clin i cal team; at ten tion to cli ent pref er - ences; com pre hen sive, con tin u ous, work-based as sess ment; and time-un lim ited sup port. The em pir i cal ba sis for these prin ci ples has also been ad dressed (14). In ad di tion to these prin ci ples, screen ing for work-readi ness is avoided, so that all cli ents who ex press a de sire for work may have that op por tu - nity (15). Var i ous studies of SE in volv ing per sons with se vere mental ill ness were car ried out in the US during the 1990s, sev eral of them rig or ous in de sign. They have con sis tently found that SE yields much higher rates of com pet i tive em ploy ment than do var i ous al ter na tive strat e gies in volv ing more grad ual ap - proaches (6,16 21). A recent metaanalysis sum ma riz ing 5 well-con ducted tri als com par ing SE with vo ca tional training con cludes that sub jects in SE pro grams are more likely to be in com pet i tive em ploy ment than are those re ceiv ing prevocational train ing, up to 18 months (for ex am ple, 34% vs 12% at 12 months); in that re view, the dif fer ence at 24 months, based on 2 tri als only, fa vours SE but is mar gin ally sig nif i cant (7). Fur ther, ev i dence that pro grams which score higher on a SE fidelity scale (22) achieve higher com pet i tive em ploy ment rates has also re cently been re ported (23). Sur pris ingly, these studies have gen er ally not re ported that SE ser vices have any ef fects on symp toms, self-es teem, or qual ity of life. Re cent ev i dence sug gests, how ever, that there is an as so ci a tion be tween the ex pe ri ence of work ing and im - prove ments in symp toms; in sat is fac tion with vo ca tional ser - vices, lei sure, and fi nances; and in self-es teem (3,24). Being of fered SE ser vices does not in prac tice guarantee steady em - ploy ment: as in di cated above, Crow ther and oth ers cal cu lated a mean em ploy ment rate of 34% at 12 months (7). Thus, the ac tual ef fects of work ing on these do mains could be di luted by the pres ence of many sub jects who work lit tle or not at all, even though they are re ceiv ing SE ser vices. Whether work it - self does af fect these do mains, or whether some other causal mech a nism is op er at ing, re mains un clear at pres ent. None the less, there ex ists a widely shared be lief that in te gra - tion of those with se vere mental ill ness into reg u lar set tings in the com mu nity is pref er a ble to main tain ing them in seg re - gated settings. On this ground, the finding that SE more ef fec - tively pro motes in te gra tion into reg u lar work set tings, with at worst no dis cern ible neg a tive out come for cli ents, sug gests that more SE pro grams should be de vel oped. The prac ti cal ques tion then arises as to the eco nomic fea si bil - ity of im ple ment ing SE ser vices for per sons with se vere men tal ill ness. This ar ti cle re views the ev i dence on the is sue. Pre vi ous re views have been con ducted (25 29). Most have in volved peo ple with de vel op men tal dis abil i ties. This re view in cor po rates more recent stud ies (20,21). Methods Po ten tially rel e vant studies were iden ti fied through searching the lit er a ture, iden ti fy ing ref er ences noted in lit er a ture re - views, and com mu ni cat ing with ex perts. Studies were sought that 1) de scribed SE pro grams set up in par al lel with other com mu nity sup port ser vices; 2) were de signed to es ti mate the SE pro gram s ef fects in isolation from those of other commu - nity sup port ser vices; 3) re ported at least some eco nomic data other than per cent age or time com pet i tively em ployed; and 4) tar geted a broad cross-sec tion of per sons with se vere men - tal illness. This led to ex clud ing stud ies on sev eral cri te ria: first, stud ies were ex cluded that dealt pri mar ily or wholly with SE for per - sons with de vel op men tal or phys i cal dis abil i ties, be cause per sons with se vere men tal ill ness have lower base line rates of em ploy ment and in te grat ing them into the reg u lar workforce pres ents dis tinct chal lenges. Sec ond, studies were ex cluded that re ported on in te grat ing vo ca tional re ha bil i ta - tion ei ther into ACT pro grams (30,31) or into com pre hen sive pro grams other than ACT (32), be cause the sep a rate ef fects of the vo ca tional com po nent on eco nomic outcomes cannot be iso lated. Fur ther, ACT pro grams target only a small sub group of per sons with se vere mental ill ness, es ti mated at 0.07% to 0.1% of the gen eral pop u la tion (33), whereas per sons with se - vere mental ill ness rep re sent about 2.6% of the gen eral popu - la tion (34). The eco nomic im pli ca tions of in clud ing a fo cus on vo ca tional re ha bil i ta tion in ACT pro gram ming are not dis - cussed ex plic itly in this re view. Third, stud ies that re ported only clin i cal out comes and workforce par tic i pa tion rates, with no data on earn ings, ser vice use, or service costs, were also ex cluded (for ex am ple, [6] and some pre lim i nary re sults of on go ing studies cited in [13]). As a re sult, in ci den tally, this meant that the only Ca na dian study iden ti fied in this search, car ried out in Brit ish Co lum bia (35), was ex cluded. Finally, 2 un pub lished re ports and 1 un pub lished Mas ter s the sis, cited in (25), as well as an other un pub lished re port cited in (13), which might not have been ex cluded ac cord ing to the pre vi - ous cri te ria, were not ob tained in time for in clu sion in this re view. To in crease com pa ra bil ity across stud ies, where nec es sary, data were re-ex pressed in terms of units per cli ent per year (for ex am ple, dol lars per client per year). In one case, stan - dard errors of the mean were con verted into stan dard de vi a - tions for the same rea son (20). Costs are re ported in the same units as in the orig i nal study.

3 498 The Canadian Journal of Psychiatry Vol 46, No 6 Table 1. Description of supported employment (SE) studies reviewed Study and location % a Previous work experience (entire sample) Groups (n) Duration Fidelity of SE program Nonrandomized studies Rogers and oth ers (16), Bos ton, MA 37 Mean of 12.4 months full-time in previous 5 years SE (19) b 1 year Low to medium c Drake and oth ers (18); Clark and others (38), Ru ral New Hampshire % in paid competitive employment at study start SE (71) DT (112) d,e 1½ years High Drake and oth ers (19); Clark and others (38), Ru ral New Hampshire % in paid competitive employment at study start SE (112) f 2 years High Bai ley and oth ers (50) 77.4 g 14.6% in competitive employment at study start SE (31) CSP (31) Becker and oth ers (21) % employed in any competitive job in previous 5 years SE (73) DT (41) 1 year High 2 years High Randomized studies Bond and oth ers (17, 39), In di an apo lis, IN and area 66 70% employed continuously 1 year or longer in competitive job SE (43) GE (43) 1 year High at one site, medium at the other h Drake and oth ers (40); Clark and others (41), New Hamp shire cit ies 46.9 Relatively good employment histories ; 0% in competitive employment at study start IPS (74) GST (69) 1½ years High at one site, low to medium at the other i Drake and oth ers (20), Wash ing ton, DC months paid work in previous 5 years SE (76) EVR (76) 1½ years High CSP = com mu nity sup port pro gram; DT = day treatment; EVR = en hanced vo ca tional rehabilit ation; GE = grad ual en try; GST = group skills training a Per centage with schizo phre nia or schi zoaf fec tive dis or der in en tire sam ple unless oth er wise speci fied. b Serv ices prior to in tro duc tion of SE are not clearly speci fied; they in clude day treatment and vo ca tional re ha bili ta tion (see Ta ble 2). c Some screening for work readi ness ap pears likely. A struc tured 7- week pre em ploy ment phase was in tended to pre cede ac tive job search (51). Vo ca tional re ha bili ta tion staff did not ap pear to work closely with the clini cal team. d DT in clud es skills training groups, so ciali za tion groups, and shel tered work within the men tal health cen tre; CSP in volves case man age ment with no day treat ment; GE clients had to re ceive at least 4 months of pre vo ca tional prepa ra tion be fore be com ing eli gi ble for the SE pro gram; GST in volves pre em ploy ment training to choose, get, keep a job, fol - lowed by in di vidu al ized sup port; EVR in volves sev eral agencies en dors ing the goal of com peti tive em ploy ment but re ly ing on paid work- adjustment training in shel tered work - shop set tings. En hanced re fers to the presence of a vo ca tional coun sellor who fa cili tated cli ent in volve ment with agencies. e Quasi- experimental de sign with com pari son group. f In this study, the day treat ment pro gram that served as com pari son group in the Drake and oth ers (18) study con verted to In di vid ual Place ment and Sup port (IPS). The study com pares out comes be fore and af ter this con ver sion. g Per cent age of sam ple with a psy chotic dis or der. h Sepa rate vo ca tional and mental health pro grams at one site. i At one site, em ploy ment spe cial ists em pha sized shel tered work. The in for ma tion is syn the sized in nar ra tive form. Metaanalytic methods were not used, due to im por tant differ - ences in the pro grams being com pared, and in other factors, which imply that for any given di men sion, such as em ploy - ment earnings, there is un likely to be in fact a sin gle un der ly - ing true ef fect size to be es ti mated. Con trolling for dif fer ences in pro grams, cli en teles, and other fac tors, using mul ti ple re gres sion meth ods as in (36), was not pos si ble, due to the small num ber of avail able stud ies. Results Ta ble 1 de scribes the studies in cluded in this re view. Five nonrandomized and 3 ran dom ized stud ies were iden ti fied. Fi - del ity to the SE model is gen er ally high, with the ex cep tion of the study by Rog ers and oth ers, while com par i son ser vices range from day treatment with rel a tively little em pha sis on vo ca tional re ha bil i ta tion (the com par i son pro gram in [18]) to state-of-the-art step wise pro grams. Data on em ploy ment his tory the single base line char ac ter is tic that con sis tently pre dicts em ploy ment out comes (18,20,21) are not re ported con sis tently, and it is dif fi cult to in fer to what ex tent this im - por tant char ac ter is tic ac tu ally var ies across stud ies. Ta ble 2 sum ma rizes the ef fects of SE on use of the fol low ing ser vices: hos pi tal iza tions, emer gency room (ER) visits or cri - sis ser vices use (high lighted as a ba rom e ter of in sta bil ity and likely sub se quent re source use), and a gen eral other cat e - gory that mostly in cludes case man age ment and out pa tient costs. In gen eral, con sis tent with other re ports (for ex am ple, [13]), SE does not ap pear to af fect hos pi tal iza tion rates ma te - ri ally or sig nif i cantly. The single pos si ble ex cep tion is the study by Rog ers and oth ers (16), where a re duc tion in hos pi tal days of more than 50% was ob served be tween the base line and fol low-up years. The sam ple size of 19 is very small, and

4 August 2001 Economic Impacts of Supported Employment for Persons With Severe Men tal Illness 499 Table 2. Supported employment impacts on use of services (per client per year unless otherwise specified) Study Hospitalization Emergency room or crisis services use Other service use Rogers and others (16) Drake and others (18); Clark and others (38) Drake and others (19); Clark and others (38) Bailey and others (50) Becker and others (21) Before SE: 6.5 days After SE: 2.9 days (not tested) 28.2% of IPS group rehospitalized during follow-up period; 25% of comparison group (not significant) (18). Mean length of stay for sample of n = 27 regular day treatment users at experimental site, from 2.63 to 2.00 (not significant); comparison site (n = 31 regular users), from to (not significant) (38) 25% hospitalized during baseline period (that is, follow-up period for comparison group of Drake and others [18]), 14.3% during follow-up period (P < 0.025) (19) for n = 31 regular users, from to (not significant) (38) Days of hospitalizations did not change (no numbers reported) SE: from 16.4% admitted in past year at baseline to 9.6% at follow-up. DT: from 26.8% to 22% (program by time interaction not significant) Crisis services (hours): before, 1.5 after, 0.2 (not tested) Crisis services: experimental site, 1.24 (SD 2.07) to 1.41 (SD 4.09) mean hours per client; comparison site: 1.71 (SD 3.83) to 1.67 (SD 3.4) hours From 1.67 (SD 3.40) hours to 1.86 hours (SD 3.44) (not significant) Days of crisis housing did not change (no numbers reported) Day treatment (days): before, 9.2; after, 3.5 Supported residential (days): before, 33.1, after, 5.5 Other services (hours): before, 141; after, a (no differences tested) Experimental site (community services crisis services, case management, outpatient, partial hospitalization, vocational): (SD ) to (SD 69.02) Control site: (SD ) to (SD ) (site by time interaction not significant at 0.05 level: P < 0.09) From (SD ) hours to (SD ) (not tested) Outpatient mental health service use did not change (no numbers reported) Bond and others (17,39) Accelerated: 5.8 days. Gradual: 5.3 days (not significant) Accelerated: 85.8 employment specialist hours (n = 25). Gradual: 40.8 hours (not tested) Drake and others (19); Clark and others (41) Similar large decreases for both groups (numbers not specified) Drake and others (20) SE: 20.9 (baseline rate 30.3). EVR: 12.1 (baseline rate 17.4) (difference not significant when adjusted for difference in baseline rate) Over 18 months: SE: 61.6 (SD 37.1) direct contact hours. GST: 74.1 (SD 59.2) (difference not significant) a In de creas ing or der of im por tance at base line: ther apy or medi ca tion evalua tion, non- SE vo ca tional serv ices, case man age ment, so cial or rec rea tional ac tivi ties, sup ported educ a - tion. Table 3. Supported employment impacts on vocational rehabilitation and health and social services costs (on annualized, per client basis) a Study Alternative program SE program Inpatient costs Other service costs Overall increase (decrease) or difference (SE other) in service costs Rogers and others (16) $620 $7128 new program + $607 usual vocational services Before: $2220 After: $1011 (not tested) Before: $8319 After: $5053 (not tested) $2639 (not tested) Drake and others (18); Clark and others (38) SE site baseline: $7686 b $1595 (SD$ 1261) SE: before, $1130 (SD$ 3227); after, $864 (SD$ 2476) (not significant) DT: before, $9300 (SD$ ); after, $ (SD$ ) (not significant) SE: before, $6349; after, $8089 DT: before, $6325; after, $5897 c SE: ($5670) DT: ($482) Drake and others (19); Clark and others (38) $6597 (SD$ 8315) $1878 (SD$ 2003) Before: $ (SD$ ) After: $ (SD$ ) Before: $5897 After: $7958 c ($2080) (not significant) d Bailey and others (50) Service costs did not change (no numbers reported) Bond and others (17,39) $6103 ($4667 day treatment + $1436 SE) $4463 ($1443 day treatment + $3020 SE) Accelerated entry: $1713 Gradual entry: $2371 e (not tested) ($2298) (not tested) Drake and others (40); Clark and others (41) $3757 $3688 SE: $4095 GST: $4457 (not significant) Outpatient costs: SE: $5525 GST: $6685 (not significant) ($1453) (not significant) a Becker and oth ers (21) and Drake and oth ers (20) do not re port any cost data and are not in cluded in this table. b Weighted av er age of both sites at base line and com pari son site at follow- up. Since this aver age was cal cu lated by the author, no stan dard deviation or test is re ported. c Case man age ment, out pa tient, emer gency, and (af ter pe riod in SE) par tial hos pi tali za tion costs. d Count ing only com mu nity costs, dif fer ence is ($2657) and is sta tis ti cally sig nifi cant using a 1- tailed test (P < 0.05). e Other serv ice costs include, in de creas ing order of im por tance: drop- in cen tre, outpatient serv ices, medi ca tion clinic, club house, psy chia trist, sub stance abuse coun sel ling.

5 500 The Canadian Journal of Psychiatry Vol 46, No 6 Table 4. Supported employment impacts on earnings, benefits, and taxes paid (per client per year) Study Employment Earnings Benefits Taxes Rogers and others (16) Increase in earnings of $1846 (not tested) Reduction in transfer payments of $933 (not tested) Increase in taxes of $425 (not tested) Drake and others (18,19); Clark and others (38) Statistically significant difference in favour of SE, counting all clients (amounts not reported) Bailey and others (50) Significant improvement in total wages for SE group (P = 0.001) (amounts not reported) Becker and others (21) SE: $340. DT: $91 (not significant) Bond and others (17,39) Accelerated entry: $1525 Gradual entry: $574 (not tested) Drake and others (40); Clark and others (41) SE: $2897 (3882). GST: $1783 (1918) SE: $6661. GST: $6912 (not significant) SE: $471. GST: $277 (P < 0.05) (P < 0.05) a Drake and others (20) SE: $1333 (2254). EVR: $1337 (1967) (not significant) a Dif fer ence is sta tis ti cally sig nifi cant as shown when pre vi ous work is used as a covaria te (40). no con fi dence in ter val or sta tis ti cal tests are re ported, so this find ing could be the re sult of sam pling vari a tion. If it is real, how ever, the fact that em ploy ment spe cial ists worked very in ten sively with cli ents (ra tio 10:1), and to some ex tent as - sumed a case-man age ment role, could have meant that they had the same kind of ef fect on hos pi tal iza tion rates as is usual with in ten sive case man age ment or ACT (36,37). There are fewer re sults on ER vis its or cri sis ser vices than on hos pi tal iza tions, but those avail able sug gest the same pattern as found with hos pi tal iza tions: no ef fect, or per haps an ef fect if em ploy ment spe cial ists also serve as case man ag ers. In terms of other ser vice use, Rog ers and oth ers (16) at trib ute an in crease in sup ported ed u ca tion hours to 2 cli ents who re - turned to school. They also at trib ute a sig nif i cant in crease in use of so cial and rec re ational ac tiv i ties to strong efforts by pro gram staff to help cli ents strengthen their so cial sup port net works. In some of the New Hamp shire day treat ment con - ver sion studies (18,19,38), the ap par ent re duc tion in commu - nity-care hours fol low ing con ver sion of day treat ment into SE is due to a dra matic re duc tion in partial hos pi tal iza tion (clo sure of day treat ment) that more than off sets smaller in - creases in vo ca tional, out pa tient coun sel ling, and case-man - age ment hours. (As re flected in the cost data pre sented be low, how ever, the day treat ment ser vices were de liv ered in group for mat and cost less per cli ent per hour than the more in ten - sive SE ser vices that re placed them.) Finally, in the study by Bond and oth ers, con trol group sub jects could re ceive SE, but only af ter a de lay of 4 months or more, dur ing which they re - ceived prevocational prep a ra tion (17,39). The em ploy - ment-spe cial ist hours are higher in the SE group. Ta ble 3 brings to gether find ings on vo ca tional re ha bil i ta tion costs and health care and so cial ser vices costs. The im pact of SE on over all costs of vo ca tional re ha bil i ta tion var ies dra mat - i cally from one study to the next. At one ex treme, in the study by Rog ers and oth ers (16), a high-in ten sity form of SE was sim ply added to ex ist ing ser vices in a set ting where vo ca - tional re ha bil i ta tion ser vices were min i mal. At the other ex - treme, in the New Hamp shire day treat ment con ver sion stud ies for which de tailed data are available (18,19,38), an ex pen sive day treat ment pro gram was closed and re placed with a less costly SE pro gram. In be tween, in the New Hamp - shire ran dom ized study (40,41), 2 sim i larly ex pen sive vo ca - tional pro grams were com pared. The study by Bond and oth ers (17,39) pres ents yet a dif fer ent in ter me di ate case, where cli ents in the ac cel er ated-en try con di tion spent less time in day treat ment, re sult ing in some what lower overall net costs for vo ca tional re ha bil i ta tion services. In pa tient costs closely re flect find ings on hos pi tal days re - ported above: the eco nomic sig nif i cance of any SE ef fects on hos pi tal iza tions ap pears neg li gi ble, ex cept in the pos si ble case of the Rog ers and oth ers study (16). The pattern ob served for other service costs is con sis tent with the above ob ser va tions. In the study by Rog ers and oth ers (16), the costs of other ser vices are re duced be cause the SE spe cial ists sub sti tute for part of those ser vices. In the New Hamp shire day treat ment con ver sion stud ies for which we have cost data, day treat ment clo sure and re duced re ha bil i ta - tion staff ing ap pear to be as so ci ated with in creased case-man age ment and out pa tient costs. The 2 other stud ies for which such costs are avail able show some what lower other costs (by < 30% in both cases) as so ci ated with SE, due mostly to some what lower out pa tient costs.

6 August 2001 Economic Impacts of Supported Employment for Persons With Severe Men tal Illness 501 The last col umn of Ta ble 3 reports that in ev ery study ex cept the one by Rog ers and oth ers (16), SE is as so ci ated with nom - i nally lower ser vice costs, tak ing vo ca tional, in pa tient, and as sorted out pa tient costs into ac count. (In the study by Bond and oth ers [17,39] hos pi tal days were slightly fewer in the SE con di tion; had they been costed, the dif fer ence in fa vour of SE would have been slightly greater.) Ta ble 4 shows dif fer ences in em ploy ment earn ings, benefits, and taxes as so ci ated with SE. As ex pected given the greater suc cess of SE at in creas ing em ploy ment rates, em ploy ment earn ings are, in ev ery case ex cept one, much greater with SE than with the al ter na tives (though still very mod est when av - er aged over all cli ents). The ex cep tion is a ran dom ized study by Drake and oth ers (20), where total earn ings are al most iden ti cal. In that study, 94% of wages earned in the SE con di - tion came from com pet i tive em ploy ment (av er age hourly wage, USD5.82), com pared with 8% in the en hanced vo ca - tional re ha bil i ta tion (EVR) con di tion. Two-thirds of EVR wages came from shel tered em ploy ment, at an av er age hourly rate of USD1.00. Im pacts on gov ern ment ben e fits and on taxes are re ported in only 2 stud ies and ap pear neg li gi ble, con sis tent with the low av er age earn ings. None of the studies re viewed re port any mea sures of care giver time or ex pense, nor do they mea sure jus tice-re lated costs. Discussion Per haps the most striking ob ser va tion to emerge from this re - view, con sis tent with Clark s pre vi ous ob ser va tion made on the ba sis of the New Hamp shire studies alone (27), is the ex - tent to which the net ef fect on vo ca tional re ha bil i ta tion costs of pro vid ing SE ser vices de pends on the con text. In a setting where no, or hardly any, vo ca tional re ha bil i ta tion ser vices are pro vided at the out set, in tro duc ing an SE ser vice is likely to in crease vo ca tional re ha bil i ta tion costs, simply be cause there is no op por tu nity for sub sti tu tion. At the other ex treme, con - vert ing ex ist ing vo ca tional pro grams into sup ported em ploy - ment ap pears to al low a sig nif i cant cost re duc tion. The find ings sum ma rized here offer little hope for a sig nif i - cant re duc tion in other health care costs (such as in pa tient care) fol low ing the in tro duc tion of SE un less, per haps, the SE spe cial ist as sumes the role of case man ager in the ab sence of such ser vices. In that case, how ever, the em ploy ment spe cial - ist s ef fec tive ness is likely to be com pro mised: in a re cent study, ded i ca tion of SE spe cial ists to a vo ca tional role (as op - posed to their as sum ing a more gen er a list role that in cludes vo ca tional re ha bil i ta tion) has been found to be strongly as so - ci ated with the success of an SE pro gram (23). The ab sence of sig nif i cant re duc tions in other health care costs could, how ever, be partly at trib ut able to the way some of the pro grams were funded. In the New Hamp shire ran dom - ized study, SE ser vices were paid out of a re search grant, and there was lit tle in cen tive for either of the 2 men tal health cen ters to limit out pa tient treat ment costs (27). It is per haps also at trib ut able to the rel a tively short fol low-up pe riod of the studies. Thus, from the point of view of a hos pi tal or com mu nity cen - tre, fi nanc ing an SE pro gram without new funds ap pears un - likely un less the use of other vo ca tional pro grams can be cur tailed. The study by Bond and oth ers (17,39) found that this could be ac com plished with out con ver sion of existing ser vices. That find ing, how ever, may be spe cific to the state s fi nanc ing sys tem at that time, wherein pro vid ers billed Medicaid on a fee-for-ser vice ba sis for day treat ment and other ser vices. In Ca na dian set tings, hos pi tals and commu - nity cen tres op er ate out of global bud gets; there is un likely to be any off set ting re duc tion in vo ca tional re ha bil i ta tion costs un less, as in the New Hamp shire and Rhode Is land experi - ments, ex ist ing pro grams are con verted into SE pro grams. (Un der man aged care in the US, sim i lar in cen tives also apply [27]). The find ings re ported in Ta ble 4 sug gest that the cost of SE ser vices ranges roughly from USD2000 to USD4000 per cli - ent yearly, al low ing for in fla tion. Using a con ver sion factor that takes rel a tive pur chas ing power into ac count rather than cur rency ex change rates, this cor re sponds to about CAD2500 to CAD5000 per cli ent yearly. The wages paid to em ploy - ment spe cial ists, the amount of over head, and the ra tio of cli - ents to em ploy ment spe cial ists (which nor mally should range be tween 20:1 and 25:1 once the pro gram has reached ma tu - rity) are the main fac tors that would in flu ence where in that range the cost per cli ent is likely to fall. As suming a fairly pessimistic sce nario, if an em ploy ment spe cial ist costs CAD per year in clud ing ben e fits, with over head and other ex penses add ing an other 50%, and if the cli ent-to-staff ra tio is 20:1, the cost per cli ent works out to CAD4500. In many Ca na dian set tings, a lower fig ure may be achiev able. Given this rel a tively modest cost, hos pi tals that have day treat ment cen tres that do not serve as a sub sti tute for acute hos pi tal iza tion, but that have be come long-term re ha bil i ta - tion or day-care fa cil i ties, are likely to find con ver sions to SE fea si ble. The au thor has pre vi ously es ti mated the di rect costs of day pro grams (ex clud ing day hos pi tal iza tion and in clud ing shel tered em ploy ment) at 2 psy chi at ric hos pi - tals in Mon treal to be in the range of CAD110 to CAD140 daily (doc u ment avail able from the au thor). A suf fi ciently large such pro gram, serv ing an av er age of about 30

7 502 The Canadian Journal of Psychiatry Vol 46, No 6 par tic i pants each week day, could be con verted into an SE pro gram serv ing sev eral times that num ber of cli ents. The foregoing anal y sis ig nores pos si ble changes in costs over time, be gin ning with start-up costs. In the con text of a hos pi - tal or com mu nity cen tre with a global bud get, con vert ing hu - man re sources from one pro gram to an other will in cur mostly out side train ing costs and, pos si bly, some space conversion costs; the re duced ef fi ciency during the team de vel op ment pe riod will re sult in fewer ser vices be ing given but not in in - creased costs as such. Later, it is also pos si ble that some cli - ents will be come able to re main em ployed without fur ther vo ca tional sup port. Al though anecdotally this does oc cur, the ev i dence re mains lim ited and some what con tra dic tory (42,43). The foregoing anal y sis has pur posely adopted the budgetary per spec tive of a hos pi tal or com mu nity cen tre be cause this is the en tity most likely to de cide whether to im ple ment an SE pro gram. Nev er the less, other per spec tives are of in ter est as well. Among these, the one of greatest prac ti cal im por tance is that of the cli ents them selves, who will be more or less in ter ested in par tic i pat ing in SE. From their point of view, aside from the per sonal value of as sum ing a more use ful, so cially in te grated role, ef fects on health ben e fits and on total earn ings are par a - mount. Al though av er age im pacts noted in Ta ble 4 are small, for an in di vid ual cli ent the po ten tial in creases in earn ings, and con se quent loss of ben e fits, may be con sid er able. In the US, re cent fed eral leg is la tion (44) is ex pected to help many in di - vid u als with psy chi at ric dis abil i ties re turn to work while main tain ing their gov ern ment-spon sored health in sur ance (to the ex tent that states ad just their reg u la tions ac cord ingly). To date, how ever, con cerns that the in crease in earn ings will be short-lived but sufficient to trig ger a loss of ben e fits that will be dif fi cult to re es tab lish re main a major rea son why many cli ents in US set tings hes i tate to reg is ter in SE pro grams. Used to en joy ing phy si cian and hos pi tal in sur ance ben e fits that are in de pend ent of em ploy ment or wel fare sta tus, Cana - dian cli ents may have less rea son to fear in creases in earn ings than have their US coun ter parts. Uni ver sal pharmacare pro - grams avail able in some prov inces (no ta bly Que bec) fur ther re duce such con cerns. At the same time, the greater the share of their earn ings that cli ents can keep without losing their in - come-sup port ben e fits, the greater will be their in cen tive to work. In Que bec, ev ery ad di tional dol lar earned above a thresh old of CAD100 re sults in an al most equal re duc tion in in come sup port. With such a small gain in earn ings, virtually the only ben e fit to the cli ent of work ing more hours is the in - trin sic sat is fac tion of doing so. A de tailed dis cus sion of opti - mal pol icy in this re gard is beyond the scope of this pa per. Closely re lated to the cli ents per spec tive is that of their fam - ily mem bers. A sec ond ary anal y sis car ried out on the New Hamp shire day treat ment con ver sion study pro vides the only ev i dence on how fam ily mem bers view SE in the con text ad - dressed here: it suggests that fam ily re ac tions are on bal ance pos i tive (45). Data on care giver time and ex pense are needed to better char ac ter ize the costs and ben e fits of SE from the fam ily mem bers point of view. A fourth per spec tive of in ter est is that of the gov ern ment, which can ex er cise var i ous le vers to pro mote the de vel op - ment of SE pro grams. This is a difficult per spec tive to de fine. From a purely bud get ary point of view, any cost-neu tral or cost-re duc ing (from the point of view of the hos pi tal or com - mu nity cen tre) con ver sion of vo ca tional pro grams to SE is ir - rel e vant to the gov ern ment: it will have no di rect ef fect on bud gets al lo cated to hos pi tals or com mu nity cen tres. The lim - ited ev i dence re viewed here sug gests that re duc tions in wel - fare pay ments and in creases in tax rev e nues are likely to be im ma te rial from the point of view of gov ern ment decision mak ers. Fur ther, in creases in earn ings and, hence, in creases in taxes paid by, and re duc tions in ben e fits paid to, some in di - vid u als en abled to en ter the la bour force may re sult in di rectly in tax re duc tions and ben e fit in creases for other individuals, as some of a fi nite num ber of job va can cies are filled in a con - text of less-than-full em ploy ment. In other words, if an SE cli - ent finds a job, an other job-seeker may have to look lon ger and per haps stay on un em ploy ment lon ger. Such in di rect ef - fects, which will some times oc cur, will at ten u ate the already neg li gi ble bud get ary ben e fits to gov ern ment. The pos si ble im pacts of SE on jus tice-re lated costs have not been in ves ti - gated but, given the low costs for the av er age client (46), they are also likely to be im ma te rial. Thus, from a purely bud get - ary per spec tive, in the Ca na dian con text the de vel op ment of SE ser vices, fi nanced with out new money, is es sen tially ir rel - e vant to the gov ern ment (whether re gional, pro vin cial, or fed eral). Of course, no gov ern ment s per spec tive is so lim ited. The ben e fits to cli ents from in creased com mu nity in te gra tion are of some value, and it is up to con sumer ad vo cates to make that value ap pear greater to the rel e vant gov ern ment decision mak ers: for gov ern ments, in flu enced by po lit i cal con sid er - ations, there is no unique or ob jec tive way of as sign ing value to such in tan gi ble benefits. The most com pre hen sive per spec tive is that of so ci ety. This is the only com monly adopted per spec tive that is di rectly rel e - vant to no stake holder. But, be ing the most com pre hen sive, it is, econ o mists ar gue, the one that should guide re source al lo - ca tion de ci sions. In the pres ent con text, it differs from the gov ern ment per spec tive in 5 prin ci pal ways. First, it in cludes care giver costs. Sec ond, changes in costs over time be come rel e vant: as ex plained above, they may have no ef fect on a hos pi tal or com mu nity cen tre s bud get ary out lays, but they af fect the value of ser vices ac tu ally pro vided. As we have

8 August 2001 Economic Impacts of Supported Employment for Persons With Severe Men tal Illness 503 seen, little ev i dence is avail able on which to es ti mate these 2 dif fer ences. Third, earn ings, which are ir rel e vant from a gov ern ment bud - get ary per spec tive ex cept in so far as they af fect ben e fit pay - ments and tax rev e nues, need to be taken into ac count. How to do so re mains an un set tled ques tion among econ o mists. Many would ar gue that the wage paid to work ers must be equal (at least as long as the prod uct and labour mar kets are com pet i tive) to the value of what they have pro duced. Thus, wages mea sure the value of la bour force par tic i pa tion: that is the tra di tional hu man cap i tal ap proach. But at least one al ter - na tive point of view has been force fully ar tic u lated in re cent years. If the real gain from a per son s em ploy ment is mea - sured by the con se quent in crease in Gross Na tional Prod uct (GNP), and if the job filled is, as sug gested ear lier, one of a fi - nite num ber in a con text of less-than-full em ploy ment, the ac - tual ef fect on GNP of an SE cli ent taking a job may well be less than his or her wage (25). That is the basic con sid er ation un der ly ing what is called the fric tion-cost method for valuing changes in la bour force par tic i pa tion (47), a method that has been ap plied in at least one pre vi ous ar ti cle in this jour nal (48). Thus, in creases in earn ings are val ued at their nom i nal value or less, de pend ing on which of the 2 ap proaches is used. A fourth dif fer ence be tween the gov ern ment and so ci etal per - spec tives is that any changes in ben e fits or taxes paid are ir rel - e vant from a so ci etal per spec tive, ex cept in so far as they af fect ad min is tra tive ex penses: they in volve no net gain or loss to so ci ety, but merely a trans fer of re sources. Given that ef fects on av er age earn ings and changes in ben e fits and taxes ap pear min i mal, this dif fer ence, as well as the third, be tween the per - spec tive of the gov ern ment and that of so ci ety is un im por tant. Finally, a fifth dif fer ence be tween the gov ern ment and so ci - etal per spec tives is that the latter per spec tive seeks to assign value to in tan gi ble ben e fits in an ob jec tive and im par tial way, so that ben e fits ex pected from al ter na tive pro grams can be com pared. There are 3 ba sic ap proaches to do ing this. The first, and the only one used to date in anal y ses of SE pro - grams, is to mea sure ben e fits in terms of an out come ap pro - pri ate to the pro gram being in ves ti gated (for ex am ple, in the pres ent con text, weeks in com pet i tive em ploy ment). The sec - ond ap proach is to use a com mon met ric to eval u ate health care ben e fits, so that com par i sons of cost-ef fec tive ness across dif fer ent health care pro grams (for ex am ple, or tho pe - dic sur gery vs vo ca tional re ha bil i ta tion) can be made. The qual ity ad justed life year (QALY) is the best-known such met ric (49). QALYs have not of ten been used in psy chi at ric re search, and they have not, to the au thor s knowl edge, been used to value the ben e fits of SE. It is not known to what ex tent they would capture vari a tions in well-be ing as so ci ated with changes in em ploy ment sta tus among per sons with mental ill ness. The third method that could be used to value in tan gi ble ben e - fits is called will ing ness-to-pay. In con texts such as the pres - ent one, this in volves asking stake holders to in di cate how much they would be will ing to pay to achieve a cer tain ben e - fit: not only cli ents but also their fam ily mem bers, and even the pub lic at large, could be asked what value they would place on them selves, or on their loved one, or on more in di - vid u als with se vere men tal ill ness being in te grated into the workforce. Again, this has not yet been done, to the au thor s knowl edge. More re search needs to be done be fore the dif fer ence be tween the gov ern ment and so ci etal per spec tives can be fully as - sessed. Ex cept for the val u a tion of the ben e fit of in creased com mu nity in te gra tion, how ever, the dif fer ences ap pear likely to be small. In con clu sion, the ev i dence to date sug gests that conversion of day treat ment or step wise vo ca tional pro grams into SE pro grams would be cost-saving or cost-neu tral from the points of view of hos pi tals, com mu nity cen tres, and gov ern - ments and that such con ver sions are de sir able on the grounds of in creas ing the com mu nity in te gra tion of the population with se vere men tal ill ness. Avail able ev i dence fur ther sug - gests that in vest ments of new money into SE pro grams will not be ma te ri ally off set by re duc tions in other health care costs or in gov ern ment ben e fit pay ments or by in creased tax rev e nues. In the pres ent state of knowl edge, such in vest ments must there fore be mo ti vated by the value of in creas ing the com mu nity in te gra tion of per sons with men tal ill ness. These in fer ences, how ever, are based on only a few, lim ited cost-ben e fit studies and there fore remain pro vi sional. Fur - ther, be cause all the studies re viewed have been carried out in the US, and be cause there are ma te rial dif fer ences be tween the US and Can ada in ser vice fi nanc ing, in ben e fit rules, in leg is la tion af fect ing em ploy ers ac tions to ward the dis abled, and in other fac tors likely to in flu ence the costs and ben e fits of SE pro grams, the in fer ences are even more pro vi sional in the Ca na dian con text. Ad di tional in ves ti ga tions, sev eral of which are cur rently un der way in the US and Can ada, will help to clarify the costs and ben e fits of SE. Acknowledgements Sup port from the Agence d Évaluation des Technologies et Modes d Intervention en Santé (AETMIS) is grate fully acknowledged. The au thor also thanks Deborah Becker, Gary Bond, Robin Clark, Rob - ert Drake and William Gnam for their many helpful comments and sug ges tions, and Youcef Ouadahi for his assistance. References 1. Lehman AF. Vo ca tional re ha bil i ta tion in schizo phre nia. Schizophr Bull 1995; 21: VanDongen C. Qual ity of life and self-es teem in work ing and nonworking per sons with mental ill ness. Com mu nity Ment Health J 1996;32: Mueser KT, Becker DR, Torrey WC, Xie H, Bond GR, Drake RE, and oth ers. Work and nonvocational do mains of func tion ing in per sons with se vere mental ill ness: a lon gi tu di nal anal y sis. J Nerv Ment Dis 1997;185:

9 504 The Canadian Journal of Psychiatry Vol 46, No 6 4. Rosenfield S. Factors con trib ut ing to the sub jec tive qual ity of life of the chron i cally men tally ill. Jour nal of Health and So cial Be hav ior 1992;33: Jacobs H, Wissusik D, Collier R, Stackman D, Burkeman D. Cor re la tions be tween psy chi at ric dis abil i ties and vo ca tional out come. Hos pi tal and Com mu nity Psy chi a - try 1992;43: Gervey R, Bedell JR. Sup ported em ploy ment in vo ca tional re ha bil i ta tion. In: Bedell JR, ed i tor. Psy cho log i cal as sess ment and treat ment of per sons with severe men tal dis or ders Wash ing ton (DC): Tay lor and Fran cis; p Crow ther RE, Mar shall M, Bond GR, Huxley P. Helping peo ple with se vere mental ill ness to ob tain work: sys tem atic re view. BMJ 2001;322: Wehman P. Sup ported com pet i tive em ploy ment for per sons with se vere dis abil i - ties. Jour nal of Ap plied Re ha bil i ta tion Coun seling 1986;17: Drake RE. A brief his tory of the in di vid ual place ment and sup port model. Psy chi at - ric Re ha bil i ta tion Jour nal 1998;22(1): Becker D, Drake R. A working life: the in di vid ual place ment and sup port pro gram (IPS). Con cord (NH): Dartmouth Psy chi at ric Re search Cen ter; Becker D, Drake R. In di vid ual placement and sup port: a com mu nity mental health cen ter ap proach to vo ca tional re ha bil i ta tion. Com mu nity Ment Health J 1994;30: Stein LI, Santos AB. As ser tive com mu nity treat ment of persons with se vere mental ill ness. New York: WW Norton; p Bond GR, Becker DR, Drake RE, Rapp CA, Meisler N, Lehman AF, and others. Im ple menting sup ported em ploy ment as an ev i dence-based prac tice. Psychiatr Serv 2001;52: Bond G. Prin ci ples of the in di vid ual place ment and sup port model: em pir i cal sup - port. Psy chi at ric Re ha bil i ta tion Jour nal 1998;22(2): Wehman P. Sup ported em ploy ment: to ward zero ex clu sion of persons with severe dis abil i ties. In: Wehman P, Moon M, ed i tors. Vo ca tional re ha bil i ta tion and sup - ported em ploy ment. Bal ti more (MD): Paull Brookes; p Rog ers SE, Sciarappa K, Mac Don ald-wil son K, Danley K. A ben e fit-cost analysis of a sup ported em ploy ment model for per sons with psy chi at ric dis abil i ties. Eval u a - tion and Program Planning 1995;18: Bond GR, Dietzen LL, McGrew JH, Miller LD. Ac cel er ating en try into sup ported em ploy ment for per sons with se vere psy chi at ric dis abil i ties. Re ha bil i ta tion Psy - chol ogy 1995;40(2): Drake RE, Becker DR, Biesanz JC, Torrey WC, McHugo GJ, Wyzik PF. Re ha bil i - ta tive day treat ment vs sup ported em ploy ment: I. vo ca tional out comes. Com mu - nity Ment Health J 1994;30: Drake R, Becker DR, Biesanz JC, Wyzik PF, Torrey WC. Day treatment vs sup - ported em ploy ment for per sons with se vere men tal ill ness: a rep li ca tion study. Psychiatr Serv 1996;47: Drake RE, McHugo G, Bebout R, Becker D, Harris M, Bond G, and others. A ran - dom ized clin i cal trial of sup ported em ploy ment for in ner-city pa tients with severe men tal dis or ders. Arch Gen Psy chi a try 1999;56: Becker DR, Bond GR, Mc Car thy D, Thomp son D, Xie H, McHugo GJ, and others. Con verting day treat ment cen ters to sup ported em ploy ment pro grams in Rhode Is - land. Psychiatr Serv2001;52: Bond GR, Becker DR, Drake RE, Vogler KM. A fi del ity scale for the in di vid ual place ment and sup port model of sup ported em ploy ment. Re ha bil i ta tion Coun - seling Bulletin 1997;40: Becker DR, Smith J, Tanzman B, Drake RE, Tremblay T. Fi del ity of sup ported em - ploy ment pro grams and em ploy ment out comes. Psychiatr Serv 2001;52: Bond G, Resnick S, Drake R, Xie H, McHugo G, Bebout R. Does com pet i tive em - ploy ment im prove nonvocational out comes for peo ple with se vere mental ill ness? Jour nal of Clin i cal and Com mu nity Psy chol ogy. Forth com ing. 25. Clark RE, Bond GR. Costs and ben e fits of vo ca tional pro grams for peo ple with se - ri ous men tal ill ness. In: Moscarelli M, Rupp A, Sar to rius N, ed i tors. Handbook of men tal health eco nom ics and health pol icy. Sus sex (UK): John Wiley and Sons Ltd; p Rog ers SE. Cost-ben e fit stud ies in vo ca tional ser vices. Psy chi at ric Re ha bil i ta tion Jour nal 1997;20(3): Clark RE. Sup ported em ploy ment and man aged care: can they co ex ist? Psy chi at ric Re ha bil i ta tion Jour nal 1998;22(1): Clark RE, Dain BJ, Xie H, Becker DR, Drake RE. The economic ben e fits of sup - ported em ploy ment for persons with mental ill ness. The Jour nal of Mental Health Pol icy and Eco nom ics 1998;1: McCaughrin WB, Ellis WK, Rusch FR, Heal L. Cost-ef fec tive ness of sup ported em ploy ment. Mental Re tar da tion 1993;31(1): Weisbrod BA, Test MA, Stein LI. Al ter na tive to mental hos pi tal treat ment: II. eco - nomic ben e fit-cost anal y sis. Arch Gen Psy chi a try 1980;37: McFarlane WR, Dushay RA, Deakins SM, Stastny P, Lukens EP, Toran J, and oth - ers. em ploy ment out comes in fam ily-aided as ser tive com mu nity treat ment. Am J Orthopsychiatry 2000;70: Chandler D, Meisel J, Hu T-W, McGowen M, Mad i son K. A capitated model for a cross-sec tion of se verely men tally ill cli ents: em ploy ment out comes. Com mu nity Men Health J 1997;33: Conseil d Évaluation des Tech nol ogies de la Santé. Le suivi intensif en équipe dans la communauté pour personnes atteintes de trou bles mentaux graves. Montréal: Ministère de la Santé et des Ser vices sociaux; Clinical Implications Conversion of ex ist ing day treatment or less effective vocational re - habilitation programs into supported employment (SE) programs can be cost-neutral or cost-saving from a budgetary point of view and should be carried out in such cases. Where such conversions are not possible, development of SE pro - grams can be justified on the grounds that they promote community integration of persons with severe mental illness more effectively than do other methods currently available. Limi ta tions Results are based on a limited number of studies. Cost estimates from some of these studies may be significantly bi - ased. Results are based on US experience and may not generalize to Can - ada. 34. Kessler R, Berglund P, Zbao S, Leaf P, Kouzis A, Bruce M, and others. The 12-month cor re lates and prev a lence of se ri ous mental ill ness in mental health. Rockville (MD): De part ment of Health and Hu man Services Public Health Ser vice Sub stance Abuse and Mental Health Ser vices Ad min is tra tion Cen ter for Mental Health Ser vices; DHHS Pub li ca tion Num ber (SMA) Block L. The em ploy ment con nec tion: The ap pli ca tion of an in di vid ual sup ported pro gram for persons with chronic mental health prob lems. Ca na dian Jour nal of Com mu nity Mental Health 1992;11(2): Lati mer E. Economic im pacts of as ser tive com mu nity treat ment: a re view of the lit - er a ture. Can J Psy chi a try 1999;44: Mueser KT, Bond GR, Drake RE, Resnick S. Models of com mu nity care for severe men tal ill ness: A re view of re search on case man age ment. Schizophr Bull 1998;24(1): Clark RE, Bush PW, Becker DR, Drake RE. A cost-ef fec tive ness com par i son of sup ported em ploy ment and re ha bil i ta tive day treat ment. Ad min is tra tion and Pol icy in Men tal Health 1996;24(1): Bond GR, Dietzen LL, Vogler K, Katuin CH, McGrew JH, Miller LD. To ward a frame work for eval u at ing cost and ben e fits of psy chi at ric re ha bil i ta tion: 3 case ex - am ples. Jour nal of Vo ca tional Re ha bil i ta tion 1995;5: Drake RE, McHugo GJ, Becker DR, An thony WA, Clark RE. The New Hampshire study of sup ported em ploy ment for peo ple with se vere mental ill ness: vo ca tional out comes. J Con sult Clin Psychol 1996;64: Clark RE, Xie H, Becker DR, Drake RE. Ben e fits and costs of sup ported em ploy - ment from three per spec tives. The Jour nal of Be hav ioral Health Ser vices and Re - search 1998;25(1): McHugo GJ, Drake RE, Becker, DR. The du ra bil ity of sup ported em ploy ment ef - fects. Psy chi at ric Re ha bil i ta tion Jour nal 1998;22(1): Bond GR, Drake RE, Mueser KT, Becker D. An Up date on sup ported em ploy ment for peo ple with se vere men tal ill ness. Psychiatr Serv 1997;48: Golden TP, O Mara S, Ferrell C, and oth ers. A the o ret i cal con struct for benefits plan ning and as sis tance, in the ticket to work and work in cen tives im prove ment act. Jour nal of Vo ca tional Re ha bil i ta tion 2000;14: Torrey W, Becker D, Drake, R. Re ha bil i ta tive day treatment ver sus sup ported em - ploy ment: II. cli ent, fam ily, and staff re ac tions to a program change. Psychosocial Re ha bil i ta tion Jour nal 1995;18(3): Wolff N, Helminiak TW, Di a mond RJ. Es ti mated so ci etal costs of as ser tive com - mu nity mental health care. Psychiatr Serv 1995;46: Koopmanschap MA, Rutten FF, Ineveld BMv, Roijen Lv. The fric tion cost method for mea sur ing in di rect costs of dis ease. Jour nal of Health Eco nom ics 1995;14: Goeree R, O Brien BJ, Blackhouse G, Agro K, Goering P. The Val u a tion of pro - duc tiv ity costs due to pre ma ture mor tal ity: a com par i son of the hu man-cap i tal and fric tion-cost meth ods for schizo phre nia. Can J Psy chi a try 1999;44: Drummond MF, O Brien B, Stoddart GL, Torrance GW. Methods for the eco nomic eval u a tion of health care programmes, Sec ond Ed. To ronto: Ox ford Uni ver sity Press; Bailey EL, Ricketts SK, Becker DR, Xie H, Drake RE. Do long-term day treat ment cli ents ben e fit from sup ported em ploy ment? Psy chi at ric Re ha bil i ta tion Jour nal 1998;22(1): Danley KS, An thony WA. The choose-get-keep model. Amer i can Re ha bil i ta tion 1987;6 9:27 9.