- Email: [email protected]
When back pain becomes disabling: a regional analysis
33
Pain, 33 (1988) 33-39 Elsevier PA1 01193
When back pain becomes disab~~g: a regional analysis Ernest Volinn * ** * , Daniel Lai * , Steven MeKinney
* * * and John D. Loeser * 3* *
* Department of Neurological Surgery, * * Multidisciplinary Pain Center, * * * Department of Biostatistics, Uniuersity of Washington, Seattle, WA 98195 (U.S.A.) (Received 3 November 1987, accepted 9 November 1987)
Back pain is a common condition and in most cases is not disabling. We have investigated disabling back pain that Summary leads to health care utilization, time lost from work, and high costs. Disabling back pain remains of obscure origin because the focus in studying it has been too narrow. Our indicator of disability is the industrial insurance claim rate for back sprain by county (N = 39) in the State of Washington. After controlling for the size of the labor force and the proportion of workers in occupations that are particularly at risk of back sprain, we determined the effect of 3 socioeconomic factors on the claim rate: the unemployment rate, percentage receiving food stamps, and per capita income. For 2 of the 3 years studied, socioeconomic factors accounted for about one-third of the variance in the claim rate. Even though claimants of industrial insurance are employed, the unemployment rate was significantly related to the claim rate in the 3 years studied. Our interpretation is that disability is a symptom of distress. Where there is a rise in job insecurity and an attendant rise in economic insecurity. there is a greater likelihood that back pain will become disabling. Key words: Back pain; Disability;
Industrial
insurance;
Unemployment;
Low back pain, one of the most common conditions in industrial societies, is also one of the least understood. That much is to be learned about the condition is suggested by the terms employed by the medical community to describe it, among them pain of ‘obscure’ origin [21] and ‘occult’ pain [13]. The American Academy of Orthopedic Surgeons brought together a number of prominent researchers in the field for a symposium on Idiopathic Low Back Pain [31], which is another way of saying that the patient suffers from a condition whose origins remain obscure. If the origins of the condition are to be elucidated, then what is meant by the condition, that which is to be explained, must be clearly delin-
Correspondence fo: Ernest Volinn, Ph.D., Department of Neurological Surgery, RI-20, University of Washington, Seattle, WA 98195, U.S.A. 03~-3959/88/$03.50
0 1988 Elsevier Science Publishers
Stress
eated. Thus, pain and disability must be distinguished. This is a well recognized distinction [e.g., 111 but one that is not always maintained 1121.Most adults in industrial societies have occasional back pain; a Louis Harris survey reports an annual prevalence in the U.S.A. of 56% [20]. Disability is the relatively less frequent behavior that may or may not attend back pain. Back pain is here conceptualized as a precondition of disability but distinct from it. It is disability that comes to the attention of physicians and other health care practitioners, results in time lost from work, and accounts for the high costs of what is referred to as back pain. This analysis concentrates on disability. A further distinction is necessary. Back pain disability, insofar as the term is used here, excludes conditions whose signs and symptoms explain the pain: fracture, dislocation, malignancy, and penetrating injury. Only about 20% of all back conditions reported to physicians are thereby
B.V. (Biomedi~
Division)
excluded [6,19.24]. The other 80%~ remain largely unexplained. The data upon which the analysis are based are industrial insurance claims for back sprain from the State of Washington. Claims are a measure of health care utilization and time lost from work and. as such, they constitute an indicator of behavior - or back pain disability as opposed to back pain itself; claims for back sprain also exclude cases of fracture, dislocation and other conditions explained by pathophysiology. In addition, back sprain is the single most costly industrial accident. State-by-state surveys show that 19--25% of all industrial insurance claims are for back sprain and 20-42% of the costs [2,17,26,27]. Our hypothesis is that the reason why back pain disability has remained largely unexplained is that the focus in studying it has been too narrow. This stands to reason. There is an immediacy to the patient presenting with pain that induces physicians and other health care practitioners to treat symptoms rather than step back. examine the patient’s social en~ronment, and trace the manifold factors implicated in the etiology. Recently, however, the view of disability has expanded beyond the medical model. Studies have elucidated the role of factors in the patient’s immediate environment such as the workplace (31 and family [for reviews, see 22,301. We take a broader view of the problem; this approach is suggested by other studies [9,10], though none have considered industrial insurance claims for back sprain. Our aim is not to preclude factors in the patient’s immediate social environment but to step back further still and assess the effect of regional socioeconomic factors. Once regional factors are identified, more proximate factors may be incorporated into the explanation of disability. The issue here, then, is the extent to which county socioeconomic variables affect claims for back sprain.
Data and methods
Claims data are from the Washington State Department of Labor and Industries. Claimants receiving medical aid only as well as claimants
receiving both medical ard and compensation for time lost from work arc included in the study. In Washingt(~n, about 70% of the covered Labor force is insured by the state and 30% privately insured: both state and privately insured claims tire included. The dependent variable of the analysis is claim rate by county in the State of Washington. This variable consists of the ratio of number of claims for back sprain to the number in the covered labor force. The independent variables consist of 3 socioeconomic indicators for counties in the state: unemployment rate. percentage of populatiofl receiving food stamps, and per capita income (per capita wage for covered labor force in 1985). The numerator and denominator of the dependent variable. i.e.. claim rate, are from different state agencies (the former from Labor and Industries and the latter from Employment Security). However. both exclude roughly the same occupational categories, notably agricultural workers. the self-employed. and members of the armed forces. Reports on claim rates published by one state agency use denominator data from the other [28]. which sustains our confidence in the way we composed claim rates. Occupation was added to the analysis, since studies have shown that those in certain occupatmns are particularly at risk of back sprain [ 1.25,29]. Why this is so - whether back sprain is a result of physical demands of occupation or whether those who have back sprain in the first place find the occupation difficult to do -- is not known and beyond the scope of this paper. High risk occupations, including such occupations as heavy material handlers and truckers, are mainly in the broad U.S. census occupational category of operators, fabricators, laborers (OFL). Thus, the percentage of the labor force in the OFL category was added to control for physical demands of occupation. Our aim was to test the effect of socioeconomic ~~~nd~tions on the claim rate for back sprain, but socioeconomic conditions in the State of Washington are variable. The unemployment rate was two-thirds higher in 1983 than it was in 1979, and the number receiving food stamps was half again as high in 1983 as it was in 1979. Thus. it was
35
TABLE DATA
I SOURCES Source
Composition
Variable
1979
1983
7 8
7 9
10
10
1985
Claim rate % labor force operators, fabricators, laborers Unemployment
rate
% food stamps
Number receiving County population
food stamps
Per capita income
7 9
2
Per capita wage
State of Washington, Department of Labor and Industries, unpublished data. State of Washington, Department of Employment Security, unpublished data. State of Washington, Washington State Data Book, 1985, Olympia, WA, Office of Financial Management, 1986. U.S. Dept. of Commerce, 1980 Census of Population, General Population Characteristics, Washington, U.S.G.P.O., 1982. Data not available (used 1985 figures, see text for explanation). State of Washington, Occupational Projections 1985-1990, Department of Employment Security, Olympia, WA, 1986. State of Washington, Income Assistance, Community Social Service and Medical Assistance, Department of Social and Health Services, Olympia, WA, 1986. Pocket Data Book, 1979, Office of Financial Management, Olympia, WA, 1980. 8. State of Washington, Forecasts of the State and County Population by Year for Selected Age Groups: 1980-2000, Office of 9. State of Washington, Financial Management, Olympia, WA, 1986. 10 U.S. Dept. of Commerce, Local Population Estimates, West, 1984 Population and 1983 Per Capita Income Estimates for Counties and Incorporated Places, Washington, DC, U.S.G.P.O. 1. 2. 3. 4. 5. 6. 7.
necessary to collect data at different points in time. We compiled separate data sets for 1979, 1983, and 1985. The sources of the data are listed in Table I. We attempted to construct variables that were comparable across data sets, but this was not always possible. The percentage of operators, fabricators, and laborers (SOFL) was not available for 1983. However, since %OFL in 1979 accounted for 97% of the variance in OFL 1985, we used %OFL 1985 for the 1983 data set. Another variable that was unavailable was per capita income for 1985; therefore, while per capita income was used for 1979 and 1983, we used per capita wage for the covered labor force for 1985. All variables in the analysis are, either explicitly or implied, ratio variables (see Table II), and variables that take the form of a ratio pose the possibility of spurious correlation. To test this possibility, we correlated, on the one hand, the
denominator of the dependent variable (covered labor force) with, on the other, the denominator of the occupational variable (no. in labor force) and the denominators of the independent variables (number employed, civilian labor force, population); we conducted this procedure for each of the 3 data sets (1979, 1983, 1985). All correlations for each data set were 0.95 or above, which indicates that spurious correlation may be a problem. To obviate any spurious correlations we added the variable l/(covered labor force) [see 81. The regression equation constructed from these variables is: ? = a + b,X,
+ b,X,
+ b,X,
+ b,X,
+ b,X,
where ? = estimated claim rate, X, = l/(covered labor force), X2 = % operators, fabricators, laborers, X, = unemployment rate, X, = W receiving food stamps, and X, = per capita income (1979, 1983) or per capita wage (1985). All vari-
RATIO
VARIABLES
Variable
name
Dependent variable Claim rate
Occupational variable % operators, fabricators, laborers (OFL) Independent variables Unemployment rate
Ratio
Number of claims/ number in covered labor force
Number OFL/ number employed
Number unemployed/ civilian labor force
%food stamps
Number receiving stamps/population
food
per capita income (1979 and 1983 data sets)
County
per capita wage (1985data set)
County wages for covered labor force/ number in covered labor force
income/population
ables pertain to counties in the State of Washington (N = 39). In addition to this regression equation, we fit 2 other models to the data. Since one county in the state has considerably more population and claims than other counties, i.e., King County, in which Seattle is located, we tested a logarithmic model. We also tested a model in which we simply eliminated the denominators of all terms and added a variable consisting of the number in the covered labor force to control for labor force size. There were no significant changes in results between these models and the above regression model. Our discussion of results refers to the latter.
Results As Table III shows, about 2% of the labor force claim back sprain each year in the State of Washington. Of 26 states submitting claim data to the U.S. Bureau of Labor Statistics, this is the highest claim rate; it is almost 3 times the average
rate for the 26 states j17j. The Washington claim rate declined slightly from year to year. though both unemployment and percent on food stamps peak in 1983 and decline thereafter. Within the state, the range between the county with the highest rate and lowest rate is. depending on the vcar, 3.5 times or more. Results of the regression equation are shown in Table IV. Tests of significance are included in the table but are not needed because the study population was composed of the universe of all claimants of industrial back sprain in every county of the state as opposed to a sample thereof. They are noted so that this study may be compared with others. These results show that on a county level there is a clear relationship between claim rate and socioeconomic variables for all 3 years examined. After controlling for the size of the covered labor force and occupation (%OFL), the combination of socioeconomic variables accounted for successively a greater amount of variance: 15% in 1979, 30% in 1983. and 37% in 1985. For all 3 years, the combination of variables would be significant at the 0.05 level or less. The greater amount of variance explained in 1985 may be the result of the variable used for that year and not for the others, per capita wage for the covered labor force instead of per capita income.
TABLE
III
STATE OF WASHINGTON CLAIM RATE FOR SPRAIN AND SOCIOECONOMIC INDICATORS
BACK
_ Claim rate (state average) Range (39 counties) Unemployment rate (state average) % receiving food stamps (state average) Per capita income (state average) Per capita wage for covered labor force (state average)
1979
1983
1985
2.18%
1.97%
I.Xl%
1.2%-4.5%
1.2%5.1%
1.2%-4.2%
6.8%
11.2%
X.1%
5.08%
6.97%
6.35%
$8073
$9915
$18,935
37
TABLE
IV
COUNTIES IN THE STATE OF WASHINGTON CLAIM RATE FOR BACK SPRAIN
(N = 39), EFFECTS
B
l/covered
Partial
r
% operators, fabricators, laborers (OFL) % unemployment
58 food stamps
Per capita income (wage in 1985) in thousands Variance explained by unemployment, food stamps, and income (per capita wage in 1985) after controlling for covered labor force and % OFL Note: We report * P < 0.05. ** P
the unstandardized
- 186.0 (345.1)
0.00
0.158
0.46 **
(0.036) 0.174 (0.065)
0.12 *
- 0.164 (0.069)
0.15 *
- 0.022 (0.149)
0.00
B
Partial
coefficients
with standard
Of the 3 socioeconomic variables, % unemployment was entered first, since the other variables, i.e., % receiving food stamps and per capita income (wage in 1985), would be affected by it. In order to file an industrial insurance claim, employment is, of course, necessary. Yet, despite reduced employment in counties where unemployment is high, unemployment is significant for all 3 data sets and, as the regression coefficients show, the relationship is positive. Also, as would be predicted, the relationship between claim rate and income in 1983 and wage in 1985 is negative. It would follow that the relations~p between food stamps and claim rate would be positive, but regression coefficients for 1979 and 1985 are negative (the relationship is not significant in 1983). However, the negative relationships may be attributed to the large amount of variance taken up by
ON
r
E
Partial
r
3.510 (0.748)
285.9 (330.6)
0.01
- 289.0
0.05
(292.9)
0.035 (0.027) 0.113
0.20
0.044 (0.025)
0.09
0.26 *
0.136 (0.045)
0.18 *
(0.047)
15% *
VARIABLES
198.5
2.603 (1.045)
0.153 (1.345) labor force
SOCIOECONOMIC
1983
1979
Intercept
OF COUNTY
- 0.042 (0.058)
0.00
- 0.177 (0.071)
0.16 *
-0.117
0.07 *
(0.053) -0.151
0.27 **
(0.043)
30% *
37% **
errors in parentheses.
one of the other variables in each data set, occupation in 1979 and per capita wage in 1985. If these variables are removed, the relationships are insignificant, In sum, the direct relationship between claim rate and unemplo~ent and the inverse relationship between claim rate and per capita income in 1983 and per capita wage in 1985 indicate that where socioeconomic conditions are better, the claim rate is lower (or, conversely, where they are worse, the claim rate is higher).
Discussion Back pain in industrial societies is a common condition that approaches universality. The less common event is disability, or such behaviors as
health care utilization and time lost t’rom work. Since so many are subject to back pain. the minor percentage whose back pain becomes disabling comprises a major problem for society. Our analysis concerns the conditions under which what we term common back pain becomes disabling. Clearly. regional socioeconomic factors are implicated. They are not, however. determinative of disability. To fully explain the etiology, other types of factors must be considered as well. including individual factors and factors in the patient’s immediate social environment. Socioeconomic factors significantly explain variability in the rate of disability, but they do not identify exactly who will be disabled. The issue is why do socioeconomic factors affect disability rates? One explanation is that claimants, to avoid work and collect disability payments or else to anticipate a job layoff. dissemble disability - in other words. they malinger. This is an explanation that may not be dismissed because our data do not have a bearing on it. Few physicians. however. believe that many of their [IX]. In addition. patients are malingerers malingerers would receive a rewarding consequence from their act. but psychometric tests show that these are often depressed patients [23]. concerns the conseAnother explanation quences of adverse socioeconomic circumstances. particularly employment insecurity. Employment has great implications for the self [7,16], and it is not only loss of employment but its threatened loss that arouses symptoms of distress. Studies have shown that physiological indicators of distress rise among workers who. though still employed. face unemployment (systolic and diastolic blood pressure. urinary norepinephrine, serum uric acid) [5,14,15]; the seeking of psychological help also increases [4]. We have noted that even among the employed, the unemployment rate is significantly related to the disability rate in all 3 data sets. Our interpretation is that disability is another symptom of distress. Where there is a rise in job insecurity and an attendant rise in economic insecurity, there is a greater likelihood that back pain will become disabling. The implication of this for state industrial insurance systems is that individual risk factors are
I101 in themselves
sufficient
to assess
Iiiwrance
rates charged to workers and their employers. II state and regional economies are strong, claim costs to insurance systems will lessen. C‘onverselv. 11’ state and regional economics are weak additional claim costs wrill hc imposed upon state Insurance systems. The issue for health care practitioners IS ttot so much how to prevent back pain but how to prcvent back pain from becoming disabling. While drugs and other treatments based on the medical model may have short-term effects. they do not address the etiology of the problem. We have proposed that regional socioeconomic factors to ;I considerable extent exert pressure on smaller social units and individuals. Insofar as such factors contribute to disability and are neglected by health care practitioners, patients will continue to be disabled by back pain.
Acknowledgements This research was supported by a contract from the Washington State Department of Labor and Industries and a grant from the National Center for Health Services Research and Health Care Technology Assessment. We wish to thank two Washington State agencies for providing us with unpublished data, the Labor Market and Economic Analysis Division of the Employment Security Department and the Health Services Analysis Division of the Department of Labor and Industries.
References Andersson,
G.B.J., Epidemiological aspects on low-back pain in industry. Spine, 6 (1981) 53-60. Antonakes. J.A.. Claims cost of back pain. Best’s Rev.. X2 (1981). Bigos, S.J. et al.. Back injuries in Industry: a retrospective study. II. Injury factors, Spine, 11 (1980) 246-256. Catalano, R., Rook, K. and Dooley. D., Labor markets and help-seeking: a test of the employment security hypothesis. .I. Hlth Sot. Behav.. 27 (1986) 277-287. Cobb, S.. Psychologic changes in men whose Jobs were abolished, J. Psychosom. Res., 18 (1974) 245-258.
39 6 Dillane, J.F., Fry, J. and Kahon, G., Acute back syndrome: a study from general practice, Br. Med. J., 3 (1966) 82-84. 7 Erikson, K., On work and alienation, Am. Social. Rev., 51 (1986) 1-8. 8 Firebaugh, G. and Gibbs, J.P., Using ratio variables to control for population sire, Social. Meth. Res., 15 (1986) 101-117. 9 Haber, L.D., State Disability Prevalence Rates: an Ecological Analysis of Social and Economic Influences on Disability, National Institute on Disability and Rehabilitation Research, Washington, DC, 1987. 10 Howards, I., Brehm, H.P. and Nagi, S.Z., Disability: from Social Problem to Federal Program, Praeger, New York, 1980. 11 Institute of Medicine, Committee on Pain, Disability, and Chronic Illness Behavior, Pain and Disability: Clinical, Behavioral, and Public Policy Perspectives, National Academy Press, Washington, DC, 1987. 12 International Association for the Study of Pain, Subcommittee on Taxonomy, Classification of Chronic Pain: Descriptions of Chronic Pain Syndromes and Definitions of Pain Terms, Pain, Suppl. 3 (1986). 13 Johnson, A.D., A Study of Occult and Manifest Diseases, State of Washington Dept. of Labor and Industries (Mimeo), Olympia, WA, 1978. 14 Kasl, S.V. and Cobb, S., Blood pressure changes in men undergoing job loss: a preliminary report, Psychosom. Med., 32 (1970) 19-38. 15 Kasl, S.V., Cobb, S. and Brooks, G.W., Changes in serum uric acid and cholesterol levels in men undergoing job loss, JAMA, 206 (1968) 1500-1507. 16 Kelvin, P. and Jarrett, J.E., Unemployment: its Social Psychological Effects, Cambridge University Press, Cambridge, 1985. 17 Klein, B.P., Jensen, R.C. and Sanderson, L.M., Assessment of workers compensation claims for back strains/sprains, J. Occup. Med., 26 (1984) 443-448. 18 Leavitt, F. and Sweet, J.J., Characteristics and frequency of malingering among patients with low back pain, Pain, 25 (1986) 357-364.
19 Lea&t, S.S. , Johnston, T.L. and Beyer, R.D., The process of recovery: patterns in industrial back injury, Ind. Med., 40 (1971) 7-14. 20 Louis Harris and Associates, The Nuprin Pain Report, Louis Harris and Associates, New York, 1985. 21 Mankin, H.S. and Adams, R.H., Pain in the back and neck. In: R.G. Petersdorf et al. (Eds.), Principles of Internal Medicine, 10th Edn, McGraw-Hill, New York, 1983, pp. 35-45. 22 Payne, B. and Norfleet, M.A., Chronic pain and the family: a Review, Pain, 26 (1986) l-22. 23 Romano, J.M. and Turner, J.A., Chronic pain and depression: does the evidence support a relationship?, Psychol. Bull., 97 (1985) 18-34. 24 Sarno, J.E., Psychosomatic backache, J. Fam. Pratt., 5 (1977) 353-357. 25 Snook, S.H., Low back pain in industry. In: A.A. White, III and S.L. Gordon (Eds.), Symposium on Idiopathic Low Back Pain, Mosby, St. Louis, MO, 1982, pp. 23-38. 26 State of California, California Work Injuries and Illnesses, 1984, Division of Labor Statistics and Research, San Francisco, CA, 1985. 27 State of Washington, Report to the Legislature: Recommendations for Improving Washington State’s Industrial Insurance System, Joint Select Committee on Worker’s Compensation, Olympia, WA, 1985. 28 State of Washington, Washington State Work Injury and Illness Summary, 1985, Department of Labor and Industries, Industrial Insurance Division, Olympia, WA, 1987. 29 Steinberg, G.G., Epidemiology of low back pain. In: M. Stanton Hicks (Ed.), Chronic Low Back Pain, Raven Press, New York, 1982, pp. l-13. 30 Turk, D.C., Flor, H. and Rudy, T.E., Pain and families. I. Etiology, maintenance, and psychosocial impact, Pain, 30 (1987) 3-27. 31 White, III, A.A. and Gordon, S.L. (Eds.), Symposium on Idiopathic Low Back Pain, Mosby, St. Louis, MO, 1982.
Pain, 33 (1988) 33-39 Elsevier PA1 01193
When back pain becomes disab~~g: a regional analysis Ernest Volinn * ** * , Daniel Lai * , Steven MeKinney
* * * and John D. Loeser * 3* *
* Department of Neurological Surgery, * * Multidisciplinary Pain Center, * * * Department of Biostatistics, Uniuersity of Washington, Seattle, WA 98195 (U.S.A.) (Received 3 November 1987, accepted 9 November 1987)
Back pain is a common condition and in most cases is not disabling. We have investigated disabling back pain that Summary leads to health care utilization, time lost from work, and high costs. Disabling back pain remains of obscure origin because the focus in studying it has been too narrow. Our indicator of disability is the industrial insurance claim rate for back sprain by county (N = 39) in the State of Washington. After controlling for the size of the labor force and the proportion of workers in occupations that are particularly at risk of back sprain, we determined the effect of 3 socioeconomic factors on the claim rate: the unemployment rate, percentage receiving food stamps, and per capita income. For 2 of the 3 years studied, socioeconomic factors accounted for about one-third of the variance in the claim rate. Even though claimants of industrial insurance are employed, the unemployment rate was significantly related to the claim rate in the 3 years studied. Our interpretation is that disability is a symptom of distress. Where there is a rise in job insecurity and an attendant rise in economic insecurity. there is a greater likelihood that back pain will become disabling. Key words: Back pain; Disability;
Industrial
insurance;
Unemployment;
Low back pain, one of the most common conditions in industrial societies, is also one of the least understood. That much is to be learned about the condition is suggested by the terms employed by the medical community to describe it, among them pain of ‘obscure’ origin [21] and ‘occult’ pain [13]. The American Academy of Orthopedic Surgeons brought together a number of prominent researchers in the field for a symposium on Idiopathic Low Back Pain [31], which is another way of saying that the patient suffers from a condition whose origins remain obscure. If the origins of the condition are to be elucidated, then what is meant by the condition, that which is to be explained, must be clearly delin-
Correspondence fo: Ernest Volinn, Ph.D., Department of Neurological Surgery, RI-20, University of Washington, Seattle, WA 98195, U.S.A. 03~-3959/88/$03.50
0 1988 Elsevier Science Publishers
Stress
eated. Thus, pain and disability must be distinguished. This is a well recognized distinction [e.g., 111 but one that is not always maintained 1121.Most adults in industrial societies have occasional back pain; a Louis Harris survey reports an annual prevalence in the U.S.A. of 56% [20]. Disability is the relatively less frequent behavior that may or may not attend back pain. Back pain is here conceptualized as a precondition of disability but distinct from it. It is disability that comes to the attention of physicians and other health care practitioners, results in time lost from work, and accounts for the high costs of what is referred to as back pain. This analysis concentrates on disability. A further distinction is necessary. Back pain disability, insofar as the term is used here, excludes conditions whose signs and symptoms explain the pain: fracture, dislocation, malignancy, and penetrating injury. Only about 20% of all back conditions reported to physicians are thereby
B.V. (Biomedi~
Division)
excluded [6,19.24]. The other 80%~ remain largely unexplained. The data upon which the analysis are based are industrial insurance claims for back sprain from the State of Washington. Claims are a measure of health care utilization and time lost from work and. as such, they constitute an indicator of behavior - or back pain disability as opposed to back pain itself; claims for back sprain also exclude cases of fracture, dislocation and other conditions explained by pathophysiology. In addition, back sprain is the single most costly industrial accident. State-by-state surveys show that 19--25% of all industrial insurance claims are for back sprain and 20-42% of the costs [2,17,26,27]. Our hypothesis is that the reason why back pain disability has remained largely unexplained is that the focus in studying it has been too narrow. This stands to reason. There is an immediacy to the patient presenting with pain that induces physicians and other health care practitioners to treat symptoms rather than step back. examine the patient’s social en~ronment, and trace the manifold factors implicated in the etiology. Recently, however, the view of disability has expanded beyond the medical model. Studies have elucidated the role of factors in the patient’s immediate environment such as the workplace (31 and family [for reviews, see 22,301. We take a broader view of the problem; this approach is suggested by other studies [9,10], though none have considered industrial insurance claims for back sprain. Our aim is not to preclude factors in the patient’s immediate social environment but to step back further still and assess the effect of regional socioeconomic factors. Once regional factors are identified, more proximate factors may be incorporated into the explanation of disability. The issue here, then, is the extent to which county socioeconomic variables affect claims for back sprain.
Data and methods
Claims data are from the Washington State Department of Labor and Industries. Claimants receiving medical aid only as well as claimants
receiving both medical ard and compensation for time lost from work arc included in the study. In Washingt(~n, about 70% of the covered Labor force is insured by the state and 30% privately insured: both state and privately insured claims tire included. The dependent variable of the analysis is claim rate by county in the State of Washington. This variable consists of the ratio of number of claims for back sprain to the number in the covered labor force. The independent variables consist of 3 socioeconomic indicators for counties in the state: unemployment rate. percentage of populatiofl receiving food stamps, and per capita income (per capita wage for covered labor force in 1985). The numerator and denominator of the dependent variable. i.e.. claim rate, are from different state agencies (the former from Labor and Industries and the latter from Employment Security). However. both exclude roughly the same occupational categories, notably agricultural workers. the self-employed. and members of the armed forces. Reports on claim rates published by one state agency use denominator data from the other [28]. which sustains our confidence in the way we composed claim rates. Occupation was added to the analysis, since studies have shown that those in certain occupatmns are particularly at risk of back sprain [ 1.25,29]. Why this is so - whether back sprain is a result of physical demands of occupation or whether those who have back sprain in the first place find the occupation difficult to do -- is not known and beyond the scope of this paper. High risk occupations, including such occupations as heavy material handlers and truckers, are mainly in the broad U.S. census occupational category of operators, fabricators, laborers (OFL). Thus, the percentage of the labor force in the OFL category was added to control for physical demands of occupation. Our aim was to test the effect of socioeconomic ~~~nd~tions on the claim rate for back sprain, but socioeconomic conditions in the State of Washington are variable. The unemployment rate was two-thirds higher in 1983 than it was in 1979, and the number receiving food stamps was half again as high in 1983 as it was in 1979. Thus. it was
35
TABLE DATA
I SOURCES Source
Composition
Variable
1979
1983
7 8
7 9
10
10
1985
Claim rate % labor force operators, fabricators, laborers Unemployment
rate
% food stamps
Number receiving County population
food stamps
Per capita income
7 9
2
Per capita wage
State of Washington, Department of Labor and Industries, unpublished data. State of Washington, Department of Employment Security, unpublished data. State of Washington, Washington State Data Book, 1985, Olympia, WA, Office of Financial Management, 1986. U.S. Dept. of Commerce, 1980 Census of Population, General Population Characteristics, Washington, U.S.G.P.O., 1982. Data not available (used 1985 figures, see text for explanation). State of Washington, Occupational Projections 1985-1990, Department of Employment Security, Olympia, WA, 1986. State of Washington, Income Assistance, Community Social Service and Medical Assistance, Department of Social and Health Services, Olympia, WA, 1986. Pocket Data Book, 1979, Office of Financial Management, Olympia, WA, 1980. 8. State of Washington, Forecasts of the State and County Population by Year for Selected Age Groups: 1980-2000, Office of 9. State of Washington, Financial Management, Olympia, WA, 1986. 10 U.S. Dept. of Commerce, Local Population Estimates, West, 1984 Population and 1983 Per Capita Income Estimates for Counties and Incorporated Places, Washington, DC, U.S.G.P.O. 1. 2. 3. 4. 5. 6. 7.
necessary to collect data at different points in time. We compiled separate data sets for 1979, 1983, and 1985. The sources of the data are listed in Table I. We attempted to construct variables that were comparable across data sets, but this was not always possible. The percentage of operators, fabricators, and laborers (SOFL) was not available for 1983. However, since %OFL in 1979 accounted for 97% of the variance in OFL 1985, we used %OFL 1985 for the 1983 data set. Another variable that was unavailable was per capita income for 1985; therefore, while per capita income was used for 1979 and 1983, we used per capita wage for the covered labor force for 1985. All variables in the analysis are, either explicitly or implied, ratio variables (see Table II), and variables that take the form of a ratio pose the possibility of spurious correlation. To test this possibility, we correlated, on the one hand, the
denominator of the dependent variable (covered labor force) with, on the other, the denominator of the occupational variable (no. in labor force) and the denominators of the independent variables (number employed, civilian labor force, population); we conducted this procedure for each of the 3 data sets (1979, 1983, 1985). All correlations for each data set were 0.95 or above, which indicates that spurious correlation may be a problem. To obviate any spurious correlations we added the variable l/(covered labor force) [see 81. The regression equation constructed from these variables is: ? = a + b,X,
+ b,X,
+ b,X,
+ b,X,
+ b,X,
where ? = estimated claim rate, X, = l/(covered labor force), X2 = % operators, fabricators, laborers, X, = unemployment rate, X, = W receiving food stamps, and X, = per capita income (1979, 1983) or per capita wage (1985). All vari-
RATIO
VARIABLES
Variable
name
Dependent variable Claim rate
Occupational variable % operators, fabricators, laborers (OFL) Independent variables Unemployment rate
Ratio
Number of claims/ number in covered labor force
Number OFL/ number employed
Number unemployed/ civilian labor force
%food stamps
Number receiving stamps/population
food
per capita income (1979 and 1983 data sets)
County
per capita wage (1985data set)
County wages for covered labor force/ number in covered labor force
income/population
ables pertain to counties in the State of Washington (N = 39). In addition to this regression equation, we fit 2 other models to the data. Since one county in the state has considerably more population and claims than other counties, i.e., King County, in which Seattle is located, we tested a logarithmic model. We also tested a model in which we simply eliminated the denominators of all terms and added a variable consisting of the number in the covered labor force to control for labor force size. There were no significant changes in results between these models and the above regression model. Our discussion of results refers to the latter.
Results As Table III shows, about 2% of the labor force claim back sprain each year in the State of Washington. Of 26 states submitting claim data to the U.S. Bureau of Labor Statistics, this is the highest claim rate; it is almost 3 times the average
rate for the 26 states j17j. The Washington claim rate declined slightly from year to year. though both unemployment and percent on food stamps peak in 1983 and decline thereafter. Within the state, the range between the county with the highest rate and lowest rate is. depending on the vcar, 3.5 times or more. Results of the regression equation are shown in Table IV. Tests of significance are included in the table but are not needed because the study population was composed of the universe of all claimants of industrial back sprain in every county of the state as opposed to a sample thereof. They are noted so that this study may be compared with others. These results show that on a county level there is a clear relationship between claim rate and socioeconomic variables for all 3 years examined. After controlling for the size of the covered labor force and occupation (%OFL), the combination of socioeconomic variables accounted for successively a greater amount of variance: 15% in 1979, 30% in 1983. and 37% in 1985. For all 3 years, the combination of variables would be significant at the 0.05 level or less. The greater amount of variance explained in 1985 may be the result of the variable used for that year and not for the others, per capita wage for the covered labor force instead of per capita income.
TABLE
III
STATE OF WASHINGTON CLAIM RATE FOR SPRAIN AND SOCIOECONOMIC INDICATORS
BACK
_ Claim rate (state average) Range (39 counties) Unemployment rate (state average) % receiving food stamps (state average) Per capita income (state average) Per capita wage for covered labor force (state average)
1979
1983
1985
2.18%
1.97%
I.Xl%
1.2%-4.5%
1.2%5.1%
1.2%-4.2%
6.8%
11.2%
X.1%
5.08%
6.97%
6.35%
$8073
$9915
$18,935
37
TABLE
IV
COUNTIES IN THE STATE OF WASHINGTON CLAIM RATE FOR BACK SPRAIN
(N = 39), EFFECTS
B
l/covered
Partial
r
% operators, fabricators, laborers (OFL) % unemployment
58 food stamps
Per capita income (wage in 1985) in thousands Variance explained by unemployment, food stamps, and income (per capita wage in 1985) after controlling for covered labor force and % OFL Note: We report * P < 0.05. ** P
the unstandardized
- 186.0 (345.1)
0.00
0.158
0.46 **
(0.036) 0.174 (0.065)
0.12 *
- 0.164 (0.069)
0.15 *
- 0.022 (0.149)
0.00
B
Partial
coefficients
with standard
Of the 3 socioeconomic variables, % unemployment was entered first, since the other variables, i.e., % receiving food stamps and per capita income (wage in 1985), would be affected by it. In order to file an industrial insurance claim, employment is, of course, necessary. Yet, despite reduced employment in counties where unemployment is high, unemployment is significant for all 3 data sets and, as the regression coefficients show, the relationship is positive. Also, as would be predicted, the relationship between claim rate and income in 1983 and wage in 1985 is negative. It would follow that the relations~p between food stamps and claim rate would be positive, but regression coefficients for 1979 and 1985 are negative (the relationship is not significant in 1983). However, the negative relationships may be attributed to the large amount of variance taken up by
ON
r
E
Partial
r
3.510 (0.748)
285.9 (330.6)
0.01
- 289.0
0.05
(292.9)
0.035 (0.027) 0.113
0.20
0.044 (0.025)
0.09
0.26 *
0.136 (0.045)
0.18 *
(0.047)
15% *
VARIABLES
198.5
2.603 (1.045)
0.153 (1.345) labor force
SOCIOECONOMIC
1983
1979
Intercept
OF COUNTY
- 0.042 (0.058)
0.00
- 0.177 (0.071)
0.16 *
-0.117
0.07 *
(0.053) -0.151
0.27 **
(0.043)
30% *
37% **
errors in parentheses.
one of the other variables in each data set, occupation in 1979 and per capita wage in 1985. If these variables are removed, the relationships are insignificant, In sum, the direct relationship between claim rate and unemplo~ent and the inverse relationship between claim rate and per capita income in 1983 and per capita wage in 1985 indicate that where socioeconomic conditions are better, the claim rate is lower (or, conversely, where they are worse, the claim rate is higher).
Discussion Back pain in industrial societies is a common condition that approaches universality. The less common event is disability, or such behaviors as
health care utilization and time lost t’rom work. Since so many are subject to back pain. the minor percentage whose back pain becomes disabling comprises a major problem for society. Our analysis concerns the conditions under which what we term common back pain becomes disabling. Clearly. regional socioeconomic factors are implicated. They are not, however. determinative of disability. To fully explain the etiology, other types of factors must be considered as well. including individual factors and factors in the patient’s immediate social environment. Socioeconomic factors significantly explain variability in the rate of disability, but they do not identify exactly who will be disabled. The issue is why do socioeconomic factors affect disability rates? One explanation is that claimants, to avoid work and collect disability payments or else to anticipate a job layoff. dissemble disability - in other words. they malinger. This is an explanation that may not be dismissed because our data do not have a bearing on it. Few physicians. however. believe that many of their [IX]. In addition. patients are malingerers malingerers would receive a rewarding consequence from their act. but psychometric tests show that these are often depressed patients [23]. concerns the conseAnother explanation quences of adverse socioeconomic circumstances. particularly employment insecurity. Employment has great implications for the self [7,16], and it is not only loss of employment but its threatened loss that arouses symptoms of distress. Studies have shown that physiological indicators of distress rise among workers who. though still employed. face unemployment (systolic and diastolic blood pressure. urinary norepinephrine, serum uric acid) [5,14,15]; the seeking of psychological help also increases [4]. We have noted that even among the employed, the unemployment rate is significantly related to the disability rate in all 3 data sets. Our interpretation is that disability is another symptom of distress. Where there is a rise in job insecurity and an attendant rise in economic insecurity, there is a greater likelihood that back pain will become disabling. The implication of this for state industrial insurance systems is that individual risk factors are
I101 in themselves
sufficient
to assess
Iiiwrance
rates charged to workers and their employers. II state and regional economies are strong, claim costs to insurance systems will lessen. C‘onverselv. 11’ state and regional economics are weak additional claim costs wrill hc imposed upon state Insurance systems. The issue for health care practitioners IS ttot so much how to prevent back pain but how to prcvent back pain from becoming disabling. While drugs and other treatments based on the medical model may have short-term effects. they do not address the etiology of the problem. We have proposed that regional socioeconomic factors to ;I considerable extent exert pressure on smaller social units and individuals. Insofar as such factors contribute to disability and are neglected by health care practitioners, patients will continue to be disabled by back pain.
Acknowledgements This research was supported by a contract from the Washington State Department of Labor and Industries and a grant from the National Center for Health Services Research and Health Care Technology Assessment. We wish to thank two Washington State agencies for providing us with unpublished data, the Labor Market and Economic Analysis Division of the Employment Security Department and the Health Services Analysis Division of the Department of Labor and Industries.
References Andersson,
G.B.J., Epidemiological aspects on low-back pain in industry. Spine, 6 (1981) 53-60. Antonakes. J.A.. Claims cost of back pain. Best’s Rev.. X2 (1981). Bigos, S.J. et al.. Back injuries in Industry: a retrospective study. II. Injury factors, Spine, 11 (1980) 246-256. Catalano, R., Rook, K. and Dooley. D., Labor markets and help-seeking: a test of the employment security hypothesis. .I. Hlth Sot. Behav.. 27 (1986) 277-287. Cobb, S.. Psychologic changes in men whose Jobs were abolished, J. Psychosom. Res., 18 (1974) 245-258.
39 6 Dillane, J.F., Fry, J. and Kahon, G., Acute back syndrome: a study from general practice, Br. Med. J., 3 (1966) 82-84. 7 Erikson, K., On work and alienation, Am. Social. Rev., 51 (1986) 1-8. 8 Firebaugh, G. and Gibbs, J.P., Using ratio variables to control for population sire, Social. Meth. Res., 15 (1986) 101-117. 9 Haber, L.D., State Disability Prevalence Rates: an Ecological Analysis of Social and Economic Influences on Disability, National Institute on Disability and Rehabilitation Research, Washington, DC, 1987. 10 Howards, I., Brehm, H.P. and Nagi, S.Z., Disability: from Social Problem to Federal Program, Praeger, New York, 1980. 11 Institute of Medicine, Committee on Pain, Disability, and Chronic Illness Behavior, Pain and Disability: Clinical, Behavioral, and Public Policy Perspectives, National Academy Press, Washington, DC, 1987. 12 International Association for the Study of Pain, Subcommittee on Taxonomy, Classification of Chronic Pain: Descriptions of Chronic Pain Syndromes and Definitions of Pain Terms, Pain, Suppl. 3 (1986). 13 Johnson, A.D., A Study of Occult and Manifest Diseases, State of Washington Dept. of Labor and Industries (Mimeo), Olympia, WA, 1978. 14 Kasl, S.V. and Cobb, S., Blood pressure changes in men undergoing job loss: a preliminary report, Psychosom. Med., 32 (1970) 19-38. 15 Kasl, S.V., Cobb, S. and Brooks, G.W., Changes in serum uric acid and cholesterol levels in men undergoing job loss, JAMA, 206 (1968) 1500-1507. 16 Kelvin, P. and Jarrett, J.E., Unemployment: its Social Psychological Effects, Cambridge University Press, Cambridge, 1985. 17 Klein, B.P., Jensen, R.C. and Sanderson, L.M., Assessment of workers compensation claims for back strains/sprains, J. Occup. Med., 26 (1984) 443-448. 18 Leavitt, F. and Sweet, J.J., Characteristics and frequency of malingering among patients with low back pain, Pain, 25 (1986) 357-364.
19 Lea&t, S.S. , Johnston, T.L. and Beyer, R.D., The process of recovery: patterns in industrial back injury, Ind. Med., 40 (1971) 7-14. 20 Louis Harris and Associates, The Nuprin Pain Report, Louis Harris and Associates, New York, 1985. 21 Mankin, H.S. and Adams, R.H., Pain in the back and neck. In: R.G. Petersdorf et al. (Eds.), Principles of Internal Medicine, 10th Edn, McGraw-Hill, New York, 1983, pp. 35-45. 22 Payne, B. and Norfleet, M.A., Chronic pain and the family: a Review, Pain, 26 (1986) l-22. 23 Romano, J.M. and Turner, J.A., Chronic pain and depression: does the evidence support a relationship?, Psychol. Bull., 97 (1985) 18-34. 24 Sarno, J.E., Psychosomatic backache, J. Fam. Pratt., 5 (1977) 353-357. 25 Snook, S.H., Low back pain in industry. In: A.A. White, III and S.L. Gordon (Eds.), Symposium on Idiopathic Low Back Pain, Mosby, St. Louis, MO, 1982, pp. 23-38. 26 State of California, California Work Injuries and Illnesses, 1984, Division of Labor Statistics and Research, San Francisco, CA, 1985. 27 State of Washington, Report to the Legislature: Recommendations for Improving Washington State’s Industrial Insurance System, Joint Select Committee on Worker’s Compensation, Olympia, WA, 1985. 28 State of Washington, Washington State Work Injury and Illness Summary, 1985, Department of Labor and Industries, Industrial Insurance Division, Olympia, WA, 1987. 29 Steinberg, G.G., Epidemiology of low back pain. In: M. Stanton Hicks (Ed.), Chronic Low Back Pain, Raven Press, New York, 1982, pp. l-13. 30 Turk, D.C., Flor, H. and Rudy, T.E., Pain and families. I. Etiology, maintenance, and psychosocial impact, Pain, 30 (1987) 3-27. 31 White, III, A.A. and Gordon, S.L. (Eds.), Symposium on Idiopathic Low Back Pain, Mosby, St. Louis, MO, 1982.