1. Each study had to be prospective (rather than retrospective) focusing the analysis on the dynamics of interventions;
2. The cohort of subjects had to be homogeneous regarding the basic medical condition at the onset of the study. Owing to the high prevalence among social security recipients, low back disorders were chosen as the medical criterion;
3. The cohort was to be composed of persons who were employed, or were employed immediately before work incapacity and who had furthermore been work incapacitated for 3 consecutive months upon entering the cohort;
4. The study had to cover a full range of interventions and incentives applied by social security, healthcare and employment agencies;
5. These interventions were to be evaluated according to critical social security indicators, such as work resumption or receipt of benefits or services;
6. An observation period was established lasting two years from the onset of work incapacity with at least three measurements to allow for a valid evaluation. Social security agencies in six countries - Denmark, Germany, Israel, the Netherlands, Sweden and the United States - agreed both to sponsor their own national research efforts and teams as well as to cooperate in the cross-national study of Work Incapacity and Reintegration (WIR). The project was carried out under the auspices of the Advisory Board on Social Security Research of the International Social Security Association. The Research Unit of the ISSA Secretariat took up the overall coordination of the project, and AS/tri Research and Consultancy Group, Leiden (Netherlands) was responsible for the scientific coordination.

1. Work incapacity had to be caused primarily by low back pain, i.e. pain between the lower edge of the 12th rib and the gluteal folds. Persons with spinal fractures or any type of surgery within the past 12 months were excluded, as were those whose cause of back pain resulted from infectious or malignant causes.
2. Subjects had to be employed (either full or part-time) at the onset of the work incapacity and covered, through employment, by the relevant social insurance system for sickness or disability benefits.
3. Subjects had to have stopped working due to back pain 3 months before the date of entry in the cohort, and
4. Subjects had to have stopped work completely (not only for a part of working hours) and had to be fully incapacitated for their jobs. The choice of three months for inclusion in the cohort was driven by administrative reasons - it was the earliest point where all countries were able to identify clients. Demographically, the subjects had to be aged between 18 and 59 years at the moment of inclusion in the cohort and had to have some fluency in the native language. Each country was supposed to aim at having at least 300 persons to be followed up through the entire observation period of 21 months.

• Socio-demographic characteristics including age, gender, educational level, household composition, income, and occupation. Also included were work characteristics such as employment history and working conditions. The health status characteristics measured included: life style; risky habits; exact medical condition and perceived health status. The socio-psychological aspects measured were: work motivation, attitudes on work and health, coping behavior, and opinions on incentives and disincentives offered by social security regarding work resumption.
• Employment characteristics sampled the local labour market situations, the employer's company size and branch, the economic prospects of the company, absence strategies, reintegration policy, working conditions and the employer's opinion on interventions, (dis)incentives and re-employment of the client.

• Training and education (general vocational);
• Work accommodations (e.g. adaptations in workplace, transportation, working hours);
• Motivators (e.g. wage subsidies, negative sanctions);
• Assessment of work capacity/incapacity (including rehabilitation inquiry); and
• Services (e.g. job search, daily care of children). The obvious measure of output or outcome is full or partial work resumption. The client's social benefit status was also a relevant outcome, which did not necessarily coincide with work resumption. While the theoretical model was influenced by the fact that the medical, social security and labour contexts of work incapacity and the operation of interventions and (dis)incentives on reintegration differ in some respects among the countries included in the project, the study draws on the many similarities in the various countries' approaches.

• The first (T1) was measured as soon as possible after 3 months of work incapacity. It covers a range of issues and variables, including relevant retrospective data on employment, health and work incapacity in the first 3 months. The measurement was set to coincide with the creation of the cohort. The actual time of sampling T1 shows differences between countries due to national circumstances that prolonged the inclusion period into the cohort. Usually the period of time for inclusion (and for measuring T1) was between 90 and 120 days after the date subjects stopped work and went on sick leave (referred to as sicklisting), but in some countries, clients had to be included into the cohort far later due to administrative reasons. In these countries, efforts were undertaken to make clear to the client that some questions (e.g. regarding health status) referred to the situation at 90 days after the date of sicklisting. Analyses on the effects of late date of testing T1 showed very few significant deviations.
• The second (T2) point of measurement occurred approximately 1 year after the first day of work incapacity. It focused retrospectively on interventions carried out and subjects' opinions concerning incentives and disincentives between T1 and T2, as well as outcomes for T2.
• Finally, the last measurement (T3) was taken about 2 years after the start of work incapacity. It addressed the interventions in the second year and the outcomes of these and earlier interventions, as well as certain other data relative to the relevant social security context. Thus the second year focused on the consequences of interventions during the first year as well as the actions taken to prevent permanent disability.

• Continuous working: those who worked at T2 and still worked at T3;
• Late resumers: those who were not yet working at T2, but who worked at T3;
• Relapse: those who were working at T2 but not working at T3;
• Non-resumers: those neither working at T2 nor at T3. (Note: these patterns were based on measures at two points in time. It is possible that those not working at either T2 or T3 nevertheless might have worked for some time during the first or second year but had relapsed to non-working status by T2 or T3).

The policy implications from this outcome seem very clearly to underscore the importance of intervening early with effective strategies. Most cohorts achieved little gain in year two (and there were even some losses). While it was likewise important for the study to identify effective strategies, the implications are clear that delays in application of those strategies will nullify the likelihood of success. The construct of the design and methodology required that the second and third measurement points were only after one and two years so further refinement of the optimum intervention period within the first year will need to be the subject of further studies.

From a policy perspective, these findings are troubling. Although it is generally considered to be good social policy that work should pay - i.e., that working should be more lucrative than being on benefits, nevertheless, social insurance is also based on principles of equity and adequacy. Part of the underpinning of public support for pay-as-you-go systems is the notion that workers who are contributing know that if they were to become too disabled to work, the system would replace their income at a level commensurate with a livable wage, not at mere subsistence. Therefore, planners must consider whether benefit replacement rates that cause income losses in excess of 30 percent for workers who become too disabled to work are in keeping with principles of adequacy and fairness. This question is particularly significant if the system takes an all or nothing approach that does not permit partial work, thereby forcing a worker who cannot work full-time to choose no work at all and then to have to exist on a low level of benefits. [note 2, The issue of level of income replacement is further exacerbated if we take into account that there is a ceiling on the level of income covered by social insurance programs. Those previously earning above that ceiling would have an even larger drop in income.]

• Age: In all of the cohorts (with the exception of the Netherlands), the oldest group (age 55 and over) was less likely to be working at T3, in several cohorts, this result also held true for the next-oldest group (45-54). Generally, the highest work resumption rates were found not in the youngest (under age 24), but in the next two youngest groups (ages 25-44).
• Gender: In most cohorts, females had a lower work resumption rate than males.
• Educational Level: A low educational level (lower secondary level or below) was associated with lower resumption rates.
• Household Composition: Respondents who lived alone had a consistently lower than average resumption rate. For other types of household composition (i.e., living with partner only; living with children only; living with partner and children; and other), the association with work status was less consistent across the cohorts.

• In the oldest age group (age above 55), resumption rates varied from 17 percent in the German cohort to 68 percent in the Dutch cohort;
• Within the lowest educated group, rates varied from 31 percent in Denmark to 68 percent in the Netherlands;
• Among those living alone, rates ranged from 13 percent in Germany to 55 percent in Sweden.

From a policy planning perspective, the finding that the youngest (under age 24) age group did not do as well as the next younger age group (25-44) in work resumption may well be a cause of some concern. It is certainly logical that the explanation may be that the younger age cohort is likely to have less work experience to draw upon and therefore would understandably have more difficulty in finding suitable new work after developing a disabling condition. However, the end result of this difficulty portends serious consequences both for social insurance systems as well as for the individuals involved. The financial impact for social insurance systems of young workers not returning to work is twofold. First, the system would likely have to support the under 24 year-old disability beneficiary for approximately 40 years before converting to the old-age system. Secondly, since pay-as-you-go systems depend on the contributions of current workers to fund pensioners and many of the study countries already have low worker to beneficiary ratios, every young person out of the labour force is one less long-term contributor to the system. Policy planners must make intense efforts to build effective return-to-work strategies that address the particular needs of this young population.

Similarly, the findings about gender differences should be examined as to causality. Is it simply that women who develop a disabling condition do not attempt return to work as readily because they have working spouses; or they cannot manage their household duties along with work after developing a disability; or they have fewer skills or less experience than men in their same age cohort making transition to other work more problematic; or, is there greater discrimination by employers against women who develop disabilities versus men who do? None of these answers can be ascertained from the study, but the findings should stimulate further research along these lines that would help inform future return-to-work policies.

The policy implications of the findings concerning health status are at first glance not particularly surprising - the less pain and the better functioning capacity, the higher the resumption rate in all cohorts. However, far less logical are the findings that there were marked differences in resumption rates within equal levels of pain intensity and functional capacity. In addition, as mentioned, within subgroups there were huge cross-national differences.

Hence, the disparities brought to light by these findings cannot fail but to prompt new questions as to why there are such differences among cohorts and subgroups regarding work outcomes when health status is the same. The implications are that further study should be conducted to understand coping with pain and the role that personal motivation plays in pain management. Clearly too, return to work among those with higher levels of pain intensity and more limited functioning may well be influenced positively or negatively by non-medical factors such as work flexibility, employer attitudes and accommodations.

In summary, the outcomes in terms of work resumption varied widely between the six cohorts. Even though some "baseline variables" were consistently related to resumption rates in all cohorts, these universalities do not account for the crossnational differences: also within socio-demographic subgroups, resumption rates varied considerably between cohorts. Therefore, the next question was whether the cross-cohort differences in resumption rates could be ascribed to differences in the repertoire of interventions that were undertaken in the six cohorts.

The policy implications from the findings concerning medical interventions are truly surprising and worthy of further study. The Swedish cohort had the lowest frequency of surgery, yet the best surgical outcomes as far as work resumption is concerned. It would seem that they have come closer than most countries in perfecting the selection process for good surgical candidates, and/or the way it is administered so that it has a positive effect on the improvement of health. The United States cohort was five times more likely to have had back surgery than were the Swedish cohort but with no better outcomes. Given that the United States is spending several percentage points more (as a percent of Gross Domestic Product) and in some cases, twice as much on health care than any of the other countries in the WIR study, it seems reasonable to question the efficacy of health care policy that promotes surgery as a frequently used intervention for common conditions such as back problems. The latter is particularly significant considering that the study found that, in all six cohorts, subjects who had resumed work within one year had a statistically significant better back function than those who had not. Such findings argue not only for intensified medical and vocational efforts especially in the first year after onset, but more importantly for labour market policies to support keeping the job open for that period.

Finally, the number of treatments given to the subjects was significant. Even those who did not report a better health status after a year nevertheless continued to receive many treatments. While on the one hand, this approach seems reasonable as their condition remained unimproved, on the other hand, it seems striking that so many treatments and interventions could have no beneficial effect on health status. From a health policy perspective, these findings suggest strongly that more research needs to be done in the area of measuring both the effectiveness, or lack thereof, of medical interventions as well as the most efficacious timing of such interventions.

From a policy perspective, the issue of job protection status takes on even more importance when consideration is given to the findings mentioned above i.e., that subjects who had resumed work within one year had a statistically significant better back function than those who had not. Certainly quick dismissal provisions may pre-empt any possibility of return-to-work with the current employer, but the findings show that they may also result in delayed and lower rates of work resumption overall. On the other hand, the US findings showed a rather high rate of work resumption overall (62.6 percent, of which 56.4 percent was with the old employer, 37.6 percent with a new employer; and 4.7 percent self-employed) , which may be despite or because of the quick dismissal practices more prevalent among US employers. It must therefore be considered whether in the rush to dismiss and replace the ailing worker, employers might not be doing both the worker and themselves a great disservice. In the case of the worker, such an action is likely to impede and possibly prevent his or her return to work because of the need to find a job with a new employer, a situation that can be especially problematic after acquiring a disabling condition. Furthermore, quick dismissal may also be bad policy from the employer's perspective because of the loss of an experienced worker who, if given more time, might in fact be able to return to work and whose recovery might be aided by working.

The policy lesson from these findings for employers is that they should carefully consider adopting a wait-and-see attitude towards a valued worker's absence due to illness or disability, at least for the first year after the initial absence. Certainly employers should think of using temporary workers or shifting duties if necessary, rather than resorting to dismissal too soon, because there is a definite cost to hiring a replacement worker, training him or her and waiting until productivity reaches the levels of the dismissed worker. On a macro level, countries whose labour policies permit employers to exercise liberal dismissal options may inadvertently be contributing to the growth of their own disability rolls because opportunities for sick or incapacitated workers to try returning to work with accommodations or adaptations, especially during the first year after onset, may be lost.

From a policy perspective, there is a vast difference between warnings of dismissal, especially if the employer suspects that the worker is malingering, versus the actual use of dismissal. Moreover, the timing of actual dismissal is particularly critical to the return-to-work outcomes. However, there are policy implications to be considered concerning the use of warnings especially if the worker is still not able to return to work or does return to work, but is not better. On the one hand, the Swedish findings indicate that the best therapy may be work itself. On the other hand, threats of dismissal alone run the risk of setting up an adversarial relationship between employers and workers. Obviously, the ideal response relies on carrots with the use of sticks only when necessary. Threats of dismissal must be preceded by offers of rehabilitation, accommodation and adaptations and used judiciously lest they weaken the employer/worker relationship and adversely affect return-to-work outcomes.

The policy implications from these findings are quite stunning. Not only does job protection seem to play a significant role in explaining this finding, but also whether appropriate supports are in place is extremely important. The findings indicate that the Dutch system provides a variety of supports that permit the returning worker to come back gradually and to draw on the supports needed including not only the typical job accommodations such as equipment, but also adapted working hours; change of workplace; job redesign and so forth.

The Dutch reforms of the 1990's not only brought job protection and a shift of the financial burden from the public sector to the employer, but they also put the onus on employers to take an active role in the reintegration of sicklisted employees. Employers are required to have a formal sickness prevention and absenteeism policy and are responsible for creating reintegration plans for incapacitated employees and doing everything within their means to reintegrate them in the company. The employer must hire a publicly approved privatized consulting firm to assist in the implementation of a prevention and reintegration policy.

Hence, it can be seen that the Dutch situation, wherein employers are holding the job and paying the sick or injured worker's benefits also does a great deal to promote job retention through encouraging cooperation and flexibility between the employer and the worker. In contrast, in systems where the sick or injured employee is forced to leave the current position and find work with a new company, the likelihood that the new employer would agree to be flexible about work resumption in a stepwise manner or with diminished duties is exceedingly slim and, at best, probably would only be for a short duration. Labour policies that allow for and/or encourage early dismissal of sick or disabled workers such as in Denmark may be preempting the possibility of bringing the worker back to work for the previous employer. Furthermore, social insurance disability programs in countries such as the United States that require job severance with the former employer as a prerequisite to qualifying for disability benefits may unwittingly be contributing to the problems of job reintegration for workers who develop serious illnesses or impairments.

The policy implications of linking decisions about provision of rehabilitation services to the benefit payment agency cannot be overstressed. In several of the study countries, the provision of vocational rehabilitation services was an integral component of the benefit process and typically was the responsibility of the same agency that pays the benefits. In such integrated models, there is a vested financial interest of the system to return the beneficiary to work as quickly as possible. Therefore, countries that have such systems also tend to understand and practice early identification and intervention.

These early intervention strategies contrast sharply with the situation in countries where the provision of vocational rehabilitation services is independent of benefit payment and/or is the responsibility of a different agency from the one paying sickness or disability benefits. The policy implications of these structural or administrative practices can be illustrated by the situation in the United States where the Social Security Administration has no responsibility for provision of rehabilitation services. All decisions related to payment of disability benefits are made prior to and independent of any possibilities of the efficacy of rehabilitation services. The policy implications of the United States example should be obvious - the Social Security Administration can only refer its' claimants to rehabilitation, but the actual decision concerning provision of such services is the authority of another agency. If rehabilitation takes place, it does so only after the claimant has been declared permanently and totally unable to work because of a serious impairment. Moreover, as previously stated, the claimant for such benefits cannot be working at the time he or she files for disability benefits, thus obviating any possibility for work reintegration with the current employer following rehabilitation. The study outcomes for the United States were probably positively influenced beyond what they would typically be for the majority of the states in the United States because the two states used for the studyãCalifornia and New Jerseyãhave temporary disability programs that pay cash sickness benefits. In the case of New Jersey, the benefits are paid first 26 weeks and in California for the first 52 weeks. Hence the two state models operate more like their foreign counterparts in that they provide an opportunity for rehabilitation or other interventions to be tried while still maintaining the incapacitated worker in his/her current job for at least 26 weeks and possibly a full year. In summary, it is very likely that the results for the United States cohort would have been far less successful had the participants been from the other 45 states that do not provide such opportunities for reintegration back to the former job or employer.

The policy implications of the disappointing return-to-work outcomes of the German cohort are a clear warning sign, especially to the industrialized countries, that new strategies must be devised to train or retrain workers with low education or limited skills into meaningful, viable occupations. Policies must be developed that recognize the knowledge and expertise of older workers and must be crafted in ways that utilize such skills in new ways or to train younger workers. For example, older workers with low education levels might still be able to "teach" younger workers such trades as bricklaying, carpentry, cabinetry, metal work and so forth. Trade unions should play a greater role in helping to retain as trainers and mentors, older workers who develop impairments or disabilities that prevent them from actually performing the tasks of the job on a regular basis. Such policies would serve a twofold purpose of retaining older workers in the labour force as well as perpetuating valuable skills.

In addition, care must be taken regarding the policy implications inherent in some countries use of immigrant or former welfare recipient labour, lest these groups suffer similar problems.