Original Articles

By Prof. Jose Luis Alfonso-Sanchez
Corresponding Author Prof. Jose Luis Alfonso-Sanchez
Preventive Medicine and Public Health, - Spain
Submitting Author Prof. Jose Luis Alfonso-Sanchez

Epidemiology, Occupational health, Cornell medical index, Screening test, Sensitivity analysis

Alfonso-Sanchez J. Does The Cornell Medical Index Measure Real Or Perceived Occupational Health?. WebmedCentral PUBLIC HEALTH 2010;1(10):WMC00909
doi: 10.9754/journal.wmc.2010.00909
Submitted on: 12 Oct 2010 10:56:12 AM GMT
Published on: 07 Oct 2010 06:16:08 PM GMT


Background. The Cornell Medical Index (CMI) has been widely used to measure health indicators. Although its effectiveness in detecting subjective pathologies has been demonstrated, it has not been evaluated as an epidemiological screening test to detect pathologies contrasted with clinical records.

Methods. A cohort sample with a two-stage randomized procedure was selected from occupational sectors of construction, industry and services in Valencia, on the Mediterranean coast in Spain, between 2006-2007. The CMI was completed by 30,000 workers, and was later correlated with their occupational medical data 6 months later.

Results. The CMI showed high sensitivity in areas of the nervous system and sensory organs (highest sensitivity value=97.52% and predictive value of a negative test=96.6%), mental disorders, skin and connective tissue diseases (highest also accuracy) and genitourinary disorders. Specificity was high for skin and connective tissue diseases (93.4%).

Conclusions. Construction and industrial workers had a higher adjusted risk than service workers. The CMI showed adequate sensitivity and positive predictive values and could therefore be used as a screening test in some areas.


General health status measures classify patients into different health states. Several indicators have been developed to assess either a subject’s level of perceived health, or his or her ability to carry out normal daily activities (personal tasks, work, travel, etc.).

The methodological starting point of the CMI as a general health measure was the assumption that health can be characterized by a set of scores applied to 5 aspects or dimensions of health status.

Within the CMI classification system, every individual state of health can be described by a row vector Ў (x1, x2, x3… xn), where the element xi represents the score in a specific dimension [1,2].

The CMI was first developed in 1949 in a much more rigorous format, containing both somatic and mental health aspects. It has been assessed on numerous occasions and used in different epidemiological analyses [3-7], as well as in the field of Occupational Medicine [5,6]. Similarly, abbreviated versions of the CMI have been validated [7], and various studies published in Spain demonstrate its applicability in our setting [8,9].

Although these studies describe the use of the CMI to measure a subject’s perception of how healthy he or she is, no research has been done to correlate perception concerning health with objective occupational health data. That is the general question.

The study that came closest to addressing this matter was probably a paper published almost 25 years ago[10] that correlated CMI as a predictor of cardiovascular diseases. 

The main objective of this paper was to explore the use of the CMI as a screening technique, comparing the results obtained from administration of the CMI questionnaire with the data obtained from the subject’s clinical histories, used as the standard test. A secondary objective was to identify the demographic factors influencing the CMI in a typical Mediterranean community.


The study was conducted in Valencia, a region with one and a half million inhabitants on the Mediterranean coast of Spain, with a capital city of the same name. The region enjoys a benign climate, and a health feature is the typical Mediterranean diet based on vegetables, salads and olive oil. By law, all workers must have their own clinical record and must be covered by an occupational health service.

A random sequential selection method in two stages was used to draw up the sample. Firstly, a random selection of 857 companies was carried out. The sample selection process yielded the following distribution: construction 103,500 registered workers, industry 206,800, and services 576,300. The farming population of the region was excluded directly. This sector accounted for only 3.4% (31,600) of the total working population of the area in the fourth quarter of 2002.

The size of the sample of workers was determined by the following values: Z=1.96, E=0.3%, p=0.5, and q=0.5. A total of 30,000 questionnaires were completed.

Only a few temporary workers dropped out of the study. Data were collected uniformly during 2002-2003. The pathologies were grouped according to the CMI groups.

By law, the clinical occupational histories are very similar, and comprise two parts: one addresses symptoms and signs and exhibits the same distribution as the pathological groups of the CMI. The other part records diagnosis, treatment and course. We use this second part to confirm the diagnosis. “Real” health was based on physicians’ assessments recorded in the medical record. This allows prospective follow-up of pathologies and treatments. The disorders diagnosed among the workers and recorded at initial assessment were recorded along with the disorders registered one year later.

The pathological classification of the CMI was used in this study. We first obtained the CMI self-report with the informed consent of the subjects to review their clinical data, and 12 months later we checked the clinical records for confirmed new diagnoses (the gold standard). For example epilepsy, seizures (neuropsychiatric diseases), asthma, chronic bronchitis, pneumonia (respiratory diseases), angina, congestive heart failure, hypertension (cardiovascular diseases), pyelonephritis, cystitis (urological diseases), dermatitis (dermatological diseases) etc., following the International classification of diseases (ICD-9th-CM) [11].

The summarized Spanish version of the CMI was used [9]. This version is comprised of 88 questions (closed answers), which can be grouped into 12 sections.

Section A, eyes and ears, was broken down into two groups. Likewise, section J was divided into two sub-sections, one corresponding to general status (with the variables of asthenia and anorexia), and the other to the consumption of toxic agents such as smoking - high tobacco consumer (> 20 cigarettes per day) - coffee (> 3 cups per day), and alcohol (more than 40 g per day). Questions 41, 58 and 80 were deleted.

Some sociomedical data were collected from the clinical and work records of the participating employees, including sector (services, industrial, construction), age, sex, marital status, weight in kg, height in cm, place of birth (born in the Valencian Autonomous Community or elsewhere), number of people living at home (including the worker), residence (the city of Valencia, metropolitan and rural areas), occupation or the kind of work the person does on the job, and years of professional experience.

For the occupation classification, the method proposed by Domingo and Marcos based on British Registrar General Classification of occupational status[12], was used  as follows: group I managers and liberal professions, group II personnel managers, group III intermediate board members and managers, group IV skilled manual workers, and group V semi-skilled manual workers.

Group VI (clergymen and army) was eliminated. Workers with previous occupational experience in different fields were also excluded (3%).

In order to validate the questionnaire, consensual validity was measured. For the items analysis, a global-item correlation was performed, resulting in values of over 0.55. Likewise, to assess internal consistency, inter-item correlations were calculated, yielding values of over 0.51. These tests were performed for all the CMI dimensions, as well as globally. A multivariate factorial analysis was also carried out. The internal consistency of responses to the questionnaire was also assessed with Cronbach's Alpha test, and a criterion of extraction of eigenvalues over 1 was employed for the confirmatory factorial analysis and structural equation.

There were two possible procedures for model estimation, comprising either aggregated or individualized data analysis. We opted for an aggregated model. With this approach the degrees of freedom for fixing the model are determined by the empirical number of health states. The comparison of models in logistic regression was carried out by means of the feasibility ratio test, based on the stepwise elimination procedure. The dependent variable in the regression model was the presence of patology or not with the CMI answers.

The rules followed for selecting the multiple logistic regression models were the following: firstly definition of the confounding and interaction variables; secondly evaluation of the interaction terms; and thirdly evaluation of the confounding terms, i.e., of bias and imprecision. In some cases, for deciding to include the variable in the model or not, we resorted to the Wald test. Evaluation of the confounding variables was carried out based on the reference odds ratio (OR), constructing all possible models that can be formed with the candidate confounding variables, and selecting the best model among the submodels that practically show no important differences with respect to the reference model – taking into account that a more precise model was preferred, i.e., the model in which the OR confidence interval was narrowest.

We used the reliability coefficient or correlation coefficient between two sequential tests and the reliability index or reliability coefficient square root, which resulted in values of at least 0.65 and 0.48, respectively[13-15] . We performed diagnostic tests for sensitivity, specificity, positive and negative predictive value, false positive rate and false negative rate, accuracy, J index of Youden on diagnosis security, and likelihood ratio. Statistical test were the c2 tests with qualitative variables, variance analysis and a multivariate analysis using the logit model.


The selected population were 30000 workers but  the average response rate was 94.2% for the three sectors: industry 95%, construction 95% and services 93%.

The bulk of the immigrant population belonged to the construction sector (Table 1), where 56% of the workers come from outside the region of Valencia (p<0.01). In fact, differences between the sexes in the construction sector were extremely pronounced. In terms of professional categories, Group 3 (administrators and workers in public services) was the largest in the services sector.

The study of the quantitative sociomedical variables analyzed in relation to the different occupational sectors is shown in Table 2. Significant age differences were found between the three sectors. In this regard, the variable of years of professional experience, height, weight and BMI showed relevant differences among the sectors.

Answers to the CMI were classified into pathological groups (Table 3). There were statistical differences in the answer distribution of the pathological groups. A significative finding was that the workers with the highest positive proportion answers of disorders in the different sections belonged to the services sector, with a mean of 10.3% of the total, followed by industry with 7.9%, and the construction sector, 6.1% (p<0.01).

However, there were two pathological groups which accounted for approximately twice the number of affirmative answers in the construction group: articular pathologies (distribution of positive answers: 19.7% construction, 12.8 industry, 6.6% services); and dermatological pathologies.

In relation to tobacco and alcohol, construction workers showed a considerably higher consumption rate: 67.4% were smokers and 26% were heavy smokers (Table 4). The corresponding percentages were 56% and 19.5% in the services sector. 15 % of the construction population drank either moderately or heavily, whereas the corresponding percentage for the services and industrial sectors was about 7% (p<0.01).

High consumption of coffee (> 3 cups per day) was far more frequent in the services sector, with a rate of 25% - nearly double that of the other sectors (p<0.01).

With these results, the CMI proves to be a highly sensitive measuring tool (Table 5), with figures of over 97% for the nervous system and sensory organs, followed by mental (94%) and skin and connective tissue diseases (91%). In the other pathological groups, the values were around 60%.

The sensitivity of the CMI was 92.1%, while its specificity and accuracy were 69% and 89%, respectively.

When the logit study was applied to the factors related to the pathological group of the CMI survey (Table 6), the adjustments regarding R2 did not exceed 74%, but showed a relatively high prediction percentage, better than 85%, with a significant X2 test.

From the statistical point of view, the pathologies detected with the CMI were not found to be related to any of the following factors: the number of people, immigrants, or subject place of residence.

Of all the factors taken into account, age was the most important, with an impact on respiratory, cardiac, locomotive and dermatological pathologies. The construction and industrial sectors had a lower risk for cardiovascular and locomotive pathologies after adjusting by the age, BMI, years of experience and toxic consumptions.


Sociomedical variables description. Our analysis of sociomedical variables revealed that the population we studied was relatively young (x=37 +/- 10.6). This is mainly due to the incorporation of young workers to the labor market, as a result of the demographic boom that took place in Spain in the 1960s and the great immigration of recent years.

While the average body mass index (BMI) for the global population was 24.8, this figure was slightly higher in construction workers (BMI=25.4). This may be due to the fact that this collective was slightly older than the other groups. Pathological obesity, defined by BMI > 30, was not found[16].

When the factor of place of birth was taken into account, we found that a high percentage of workers (almost 40%) had come from outside Valencia. This is probably due to the high influx of immigrants from underdeveloped countries, particularly to the construction sector (56%).

The bulk of the population in our sample lived in the city of Valencia and its metropolitan area. Apart from purely demographic reasons, this circumstance can also be explained by the large number of companies established in the city and metropolitan area, and by the fact that the agricultural sector was excluded from our study.

Another finding of note was that in general, the services sector revealed the highest number of disorders or symptoms detected by the CMI, while the construction sector registered the lowest.

In turn, the construction sector registered a higher consumption of toxic substances (tobacco and coffee), with a greater percentage of smokers (67.4%, with a high proportion of heavy smokers as well: > 20 cigarettes a day, 26.4%). Almost 15% of this sector also acknowledged daily alcohol intake, compared with 7% in the other sectors (though in the latter case real consumption is higher than admitted, according to data from other studies) [17].

A possible consequence of these negative factors may be the high accident rate in the construction sector, which is the economic activity with the highest level of accidents in the Community of Valencia (17.8% of all labor accidents) – despite the fact that this sector only represents 7% of the registered working population.

Services workers nonetheless showed the lowest levels of subjective pathologies in two categories: osteoarticular pathologies [18] and dermatological pathologie. This finding is very logical, since construction workers are exposed to numerous risks[19], and is in line with the observations of other studies which outline a greater incidence of eczema on the hands in the construction sector, due to the known irritating and sensitizing effect of cement, and to exposure to numerous sensitizing agents used in rubber gloves and boots [20].

CMI. The CMI has proven to be useful in detecting emotional and neurotic disorders and somatic impairments [3, 21-22], since it allows the patient to freely express his or her symptoms, while collecting neuropsychiatric information at the same time and also measuring anger [23] 23. In fact, studies on epidemiological disease patterns indicate that changes in the production mode can increase functional pathologies. The fact that the disorders reported in the neuropsychiatric section are clearly predominant in the services sector, together with the logit analysis, further indicates that the CMI is able to detect psychopathological disorders. Furthermore, the CMI showed the highest sensitivity score for the two groups of nervous system and mental disorders. This figure was also found in another paper that studied an older population and the concordance of CMI with structured clinical assessment for the identification of physical health status 24.

Our results showed almost no correlations among the different occupations studied and the pathologies detected by the CMI, after adjusting by other factors as the age, BMI, years of experience and toxic consumptions.

The services sector is generally considered to be privileged because it is free of classical risk factors such as physical danger, air contaminants like smoke and dust, noise, the handling of toxic substances or risk of traumatisms or radial tunnel syndromes [25]. However, the working environment of the services sector does present aggressive factors that can lead first to mental problems and later to physical disorders as well. These include a sedentary lifestyle, permanent stress, etc. Such risk factors can initially produce subtle changes in perceived health, such as irritability, mood changes or disorders in social relationships and dissatisfaction. As an example, workers in the Hamburg Construction Worker Study showing low satisfaction with their work achievements had a higher prevalence of chronic lower back pain [26]. All of these alterations can trigger neurovegetative responses, which, if long lasting, will eventually bring about organic lesions such as peptic ulcers, sustained hypertension and coronary disease [27]. There were no statistical differences among the sensitivity scores related to these factors of the three sectors.

Another factor to take into account is geography, or the physical environment proper. Some studies in the Community of Valencia [28] indicate that the risk of death from mental and central nervous system pathologies is smaller in predominantly industrial areas than in agricultural zones.

The findings of this study show that the CMI proves to be adequate as a screening test. Its sensitivity is high (92.1%), and it exhibits a strong positive predictive value (95%). Very few false negatives were detected, and the overall value of the test as a predictor of health was good, particularly in the areas of the nervous system and sensory organs, genitourinary diseases, mental disorders, and skin and connective tissue diseases. The CMI can therefore serve as a very valid instrument in predicting pathologies, particularly in these areas – as was predicted 20 years ago [10] or more [29]. There was limited sensitivity in the cardiovascular and locomotive systems. In this sense, a previously published paper studied factors segregated by the hypothalamic-pituitary-adrenal axis in patients with rheumatoid arthritis, but found no correlation with the responses to the CMI health questionnaire [30].

The idea of a relationship between disease and the CMI is almost new. A previous paper established the relationship between[31] primary health care center attendance rates and CMI scores among both husbands and wives.

Limitations and biases. Several factors may introduce biases that are difficult to control in a study of this nature. The first of these is the simple use of the working population to select the sample, because this segment enjoys better health than the population as a whole and is not representative of the unemployed population. This bias, nonetheless, has not interfered with the internal validity of the study.

A final consideration is that the respondents to the questionnaire may have forgotten data, and that this may have introduced inaccuracies in terms of the pathologies reported or an underestimation of health problems. As some authors state, the answers solicited by the CMI in certain sectors may reveal the respondents’ feelings about health, rather than their actual health status [5]. Lastly, the study was carried out among members of the working population who might have been afraid of revealing the degree of their health problems. However, this study is supported in part by another previous survey, at least in the field of somatic health [32].

A diagnostic understatement is likely, due to different reasons. For example, it is possible that in some generally chronic pathologies, confirmation of the diagnosis took more than 6 months, and also in some cases the diagnosis may not have been recorded in the clinical story (healthy effect). This would have affected the specificity and sensitivity values, and probably more in some pathologies than in others. However, we need more studies to confirm this relationship to health status.


1.-Verdon ME, Siemens K. Yield of review of systems in a self-administered questionnaire. J Am Board Fam Pract 1997;10:20-7.
2.-Mor V. Cancer patients’quality of life over the disease course: lessons from the Real World. J Chronic Dis 1987; 40:535-44.
3.-Brodmann K, Erdmann AJ, Lorge I, Wolf HG. The Cornell Medical Index-Health questionnaire II. As a diagnostic instrument. J Amer Med Ass 1951;145:152-7.
4.-Abramson JH, Terespolsky L, Brook JG, Kark SL. Cornell Medical Index as a health measure in epidemiological studies. A test of the validity of health questionnaire. Br J Prev Soc Med 1965;19:103-10.
5.-Erdmann AJ, Brodmann K, Deutcherberger J, Wolf HG. Health Questionnaire Use in an Industrial Medical Departament. Indus Med and Surg 1953; 22:355-7.
6.-Erdmann AJ. Experiencies in use of self-administered health questionnaire. Arch of Indus Med 1959;19:339-44.
7.-Levan I, Arnon A, Portnoy A. Two shortened versions of the Cornell Medical Index. A new test of validity. Int J Epidemiol 1977; 6:135-44.
8.-Salvatierra V, Cuenca C, Florido J, Moreno F. Empleo de un cuestionario de síntomas, actitudes y trastorno emocional. Rev Esp Obstet Ginecol 1987;46:531-40.
9.-Senra A, Palmeiro R. Aplicación del CMI al diagnóstico de enfermos tumorales. Rev San Hig Pub 1973;47:482-7.
10.- Weaver CA, Ko YH, Alexander ER, Pao YL, Ting N. The Cornell Medical Index as a predictor of health in a prospective cardiovascular study in Taiwan. Am J Epidemiol 1980;111:113-24.
11.-International Classification of disease. Ninth Revision, Clinical Modification. C.P.H.A. Edwards Bro, Ann Arbor, Mich. 1980.
12.-Álvarez C, Alonso J, Domingo A, Regidor E. La medición de la clase social en ciencias de la salud. Barcelona: SG editores SA. 1995:80.
13.-Campbell DT, Fiske DW. Convergent and discriminant validation by the multitrait-multimethod matrix. Psychol Bull 1959;56:81-105.
14.-Swets JA. Measuring the accuracy of diagnostic systems. Science 1988;240:1.285-93.
15.- Balogh I; Orbaek P; Winkel J; Nordander C; Ohlsson K; Ektor-Andersen J.Questionnaire-based mechanical exposure indices for large population studies--reliability, internal consistency and predictive validity. Scand J Work Environ Health 2001;27:41-8.
16.-Jakicic. The role of physical activity in prevention and treatment of body weight gain in adults. J Nutr 2002;132:3826.
17.-Herrera S, Vazquez JL, Gaite L, Diez JF, Pena C, García E. Alcohol consumption in a rural area of Cantabria. Soc Psychiatry Psychiatr Epidemiol  1996;31:199-206.
18.- Luoma K; Riihimaki H; Raininko R; Luukkonen R; Lamminen A; Viikari-Juntura E. Lumbar disc degeneration in relation to occupation. Scan J Work Environ Health 1988;24:558-66.
19.-Judd L.A descriptive study of occupational skin disease. N Z Med J 1949;27:147-9.
20.-Dejobert Y. Dermatology-allergy study in occupational medicine. Rev Prat 2002;52:1415-9.
21.-Schnurr PP, Spiro A 3rd, Aldwin CM, Stuket TA. Physical symptom trajectories following trauma exposure: longitudinal findings from the normative aging study. J Nerv Ment Dis 1998;186:522-8.
22.-Leitschuh GA. How college students’ physical health relates to coping. Psychol Rep 1999;85:220-6.
23.-Matsumoto Y, Uyama O, Shimizu S, Michishita H, Mori R, Owada T, Sugita M. Do anger and aggression affect carotid atherosclerosis? Stroke 1993;24:1761-2.
24.-Pendleton N, Clague JE, Horan MA, Rabbitt PM, Jones M, Coward R, Lowe C, McInnes L. Concordance of Cornell medical index self-reports to structured clinical assessment for the identification of physical health status. Arch Gerontol Geriatr 2004;38:261-9.
25.-Roquelaure Y; Raimbeau G; Dano C, Martin YH, Pelier-Cady MC, Mechali S, Benetti F, Mariel J, Fanello S, Penneau-Fontbonne D. Occupational risk factors for radial tunnel syndrome in industrial workers. Scan J Work Environ Health 2000; 26:507-13.
26.-Latza U; Pfahlberg A; Gefeller O. Impact of repetitive manual materials handling and psychosocial work factors on the future prevalence of chronic low-back pain among construction workers. Scan J Work Environ Health 2002;28:314-23.
27.-Nakao M, Yamanaka G, Kuboki T. Mayor depression and somatic symptoms in a mind/body medicine clinic. Psychopathology 2001;34:230-5.
28.-Morales MM, Dominguez C, Llopis A. A comparative study of mortality in agricultural and industrial areas in Spain. Eur J of Epidemiol 1995;11:633-41.
29.- Bolender CL, Swoope CC, Smith DE. The Cornell Medical Index as a prognostic aid for complete denture patients. J Prosthet Dent 1969;22:20-9.
30.- Mukai E, Nagashima M, Hirano D, Yoshino S. Comparative study of symptoms and neuroendocrine-immune network mediator levels between rheumatoid arthritis patients and healthy subjects. Clin Exp Rheumatol 2000;18:585-90.
31.-Polliack MR. The relationship between Cornell Medical Index scores and attendance rates. J R Coll Gen Pract 1971;21:453–459.
32.-Sheridan CL, Mulhern M, Martin D. Validation of a self-report measure of somatic health. Psychol Rep 1998;82:679-87.

Source(s) of Funding


Competing Interests



This article has been downloaded from WebmedCentral. With our unique author driven post publication peer review, contents posted on this web portal do not undergo any prepublication peer or editorial review. It is completely the responsibility of the authors to ensure not only scientific and ethical standards of the manuscript but also its grammatical accuracy. Authors must ensure that they obtain all the necessary permissions before submitting any information that requires obtaining a consent or approval from a third party. Authors should also ensure not to submit any information which they do not have the copyright of or of which they have transferred the copyrights to a third party.
Contents on WebmedCentral are purely for biomedical researchers and scientists. They are not meant to cater to the needs of an individual patient. The web portal or any content(s) therein is neither designed to support, nor replace, the relationship that exists between a patient/site visitor and his/her physician. Your use of the WebmedCentral site and its contents is entirely at your own risk. We do not take any responsibility for any harm that you may suffer or inflict on a third person by following the contents of this website.

1 review posted so far

A Review
Posted by Dr. Manuel,PhD Hens on 10 Dec 2010 09:19:01 PM GMT

0 comments posted so far

Please use this functionality to flag objectionable, inappropriate, inaccurate, and offensive content to WebmedCentral Team and the authors.


Author Comments
0 comments posted so far


What is article Popularity?

Article popularity is calculated by considering the scores: age of the article
Popularity = (P - 1) / (T + 2)^1.5
P : points is the sum of individual scores, which includes article Views, Downloads, Reviews, Comments and their weightage

Scores   Weightage
Views Points X 1
Download Points X 2
Comment Points X 5
Review Points X 10
Points= sum(Views Points + Download Points + Comment Points + Review Points)
T : time since submission in hours.
P is subtracted by 1 to negate submitter's vote.
Age factor is (time since submission in hours plus two) to the power of 1.5.factor.

How Article Quality Works?

For each article Authors/Readers, Reviewers and WMC Editors can review/rate the articles. These ratings are used to determine Feedback Scores.

In most cases, article receive ratings in the range of 0 to 10. We calculate average of all the ratings and consider it as article quality.

Quality=Average(Authors/Readers Ratings + Reviewers Ratings + WMC Editor Ratings)