Reliability and validity of interview data on chronic diseases. The Mini-Finland Health Survey

Reliability of perceived health by sex and age

The aim of this study was to assess the test-retest reliability of a measure of perceived general health by sex and age. The study analyzed data from the nationally representative Mini-Finland Health Survey of 8000 adults aged 30 and over. The subjects were invited to attend a personal health interview and a health examination in 1978-1980. Altogether 7217 persons participated. Perceived general health was measured at the personal health interview and in the self-administered questionnaire 1-6 weeks apart. The identical questions were: how would you assess your current health? The response alternatives were good, fairly good, intermediate, fairly poor, poor and cannot say. This study showed that among men and women unweighted agreement of the 'good-intermediate-poor' categorization of perceived health was around 70% and unweighted kappa-values were around 0.5. Only in the oldest age-group (75+) reliability declined below these levels. The fair or good reliability of perceived health observed in this study gives additional confidence for using this general measure of overall health status in future research.

The necessary role of the autopsy in cardiovascular epidemiology

The autopsy rate in the United States today is remarkably low, with proportionally fewer autopsies for natural causes of death. Consequently, most cardiovascular epidemiology studies do not use autopsy data and rely on death certificates, medical records, questionnaires, and family interviews as sources of mortality information. These practices introduce a high degree of variability and uncertainty regarding cause of death. This review illustrates the necessity for increased use of autopsies in cardiovascular epidemiology by critically evaluating other measures of cardiovascular disease (CVD) incidence. We evaluated the literature regarding CVD as cause of death and conducted discussions with cardiologists, pathologists, and epidemiologists. No attempt was made for meta-analysis. This review shows the limited reliability of death certificates, medical records, and interviews as sources of mortality statistics. In addition, the autopsy's role in clearly indicating the presence of CVD is illustrated. The autopsy used in conjunction with medical records is the only reliable means for establishing cause of death from CVD. There is an urgent need to reassess the current dependence of statistical mortality data on death certificates and other inadequate sources of CVD incidence. Death certificates, in general, are inadequately monitored for quality control and appropriate administrative oversight. With an increase in the number of hospitals performing no autopsies to investigate cause of death, a uniform national autopsy database is needed.

Cumulative incidence of lumbar disc diseases leading to hospitalization up to the age of 28 years

STUDY DESIGN

A cohort study of 12,000 persons with a follow-up period ranging from birth to the age of 28 years.

OBJECTIVES

To determine the cumulative incidence of lumbar disc disease leading to hospitalization in a well-defined geographic population, with special emphasis on age of onset and differences between the genders and among disease categories.

SUMMARY OF BACKGROUND DATA

There are few published reports on low back pain in adolescence, and most of them are cross-sectional in design.

METHODS

The Finnish National Hospital Discharge Register was used to identify all possible cases of lumbar disc disease. Copies of all hospital records regarding the diagnosis and course of the disease were collected and the hospitalizations were classified into three categories: 1) confirmed herniated nucleus pulposus (HNP), 2) unconfirmed HNP (sciatica symptoms without visible herniation), and 3) other low back diseases. The patients were classified according to the highest category of diagnosis achieved by each.

RESULTS

Symptomatic low back pain leading to hospitalization first appeared around the age of 15 years, and the incidence rose more sharply from the age of 19, especially in men with other low back diseases. Men were hospitalized more than twice as often as women.

CONCLUSIONS

Lumbar disc diseases leading to hospitalization begin early.

Chronicity of back problems during working life

STUDY DESIGN

A follow-up study of a cohort of 444 patients aged 16 to 59 years who consulted with their general practitioners (GPs) in 1987-1988 for an incident episode of back pain.

OBJECTIVES

To determine the proportion of patients with back pain in whom chronic back problems develop after a follow-up of 7 years, to compare health outcomes and labor force participation of patients with and without chronic back problems and to identify determinants of chronicity.

SUMMARY OF BACKGROUND DATA

The incidence and prevalence of back pain are very high. A large proportion of the costs related to medical consumption, absence from work, and disability are probably caused by chronic back problems. It is unknown what proportion of back problems become chronic, especially after a long follow-up period, and which factors can predict chronicity.

METHODS

Data on the course of the symptoms and medical consumption from the period between 1987-1988 and 1991 were gathered retrospectively. Data on several health outcomes, including LFP, and data on some work characteristics were collected prospectively in 1991. A more extensive data set on health outcomes including psychologic status and working situation was collected in 1994.

RESULTS

Chronic back problems developed in 28% of the patients. These patients reported more pain, higher levels of medical resource consumption, worse health outcomes, and lower labor force participation. Episodes of back pain before 1987-1988, severe pain in 1991, and disability score in 1991 were positively associated with chronicity in 1994, difficulties with job performance in 1991, and frequent stooping in the subgroup of patients who held a paying job in the follow-up period.

CONCLUSIONS

Even after a follow-up of 7 years, the proportion of people with chronic back problems was high. The consequences for quality of life, labor force participation, and consumption of medical resources are clear. Further research is necessary to examine determinants and ways to prevent chronicity.

Validity and accuracy of interview and diary data on children's medical utilisation in The Netherlands

STUDY OBJECTIVE

To assess the validity and accuracy of children's medical utilisation estimates from a health interview and diary and the possible consequences for morbidity estimates. The influence of recall bias and respondent characteristics on the reporting levels was also investigated.

DESIGN

Validity study, with the medical record of the general practitioner (GP) as gold standard. In a health interview and three week diary estimates of medical utilisation of children were asked and compared with a GP's medical record.

SETTING

General community and primary care centre in the Netherlands.

PARTICIPANTS

Parents of 1,805 children and 161 GPs.

MAIN RESULTS

The sensitivity of the interview (0.84) is higher than the diary (0.72), while specificity and kappa are higher in the diary (0.96; 0.64) than in the interview (0.91; 0.5-8). Recall bias, expressed as telescoping and heaping, is present in the interview data. Prevalence estimates of all morbidity are much higher in the interview, except for skin problems. Compared with a parental diary more consultations are reported exclusively by the GP for children from ethnic minorities (OR 1.6), jobless (OR 2.3), and less educated mothers (OR 2.6).

CONCLUSIONS

Estimates of medical utilisation rates of children are critically influenced by the method of data collection used. Interviews are prone to introduce recall bias, while diaries should only be used in populations with an adequate level of literacy. It is recommended that medical records are used, as they produce most consistent estimates.

Reliability of a self-reported health questionnaire in a periodontal disease study

This study assesses the reliability of a self-reported health questionnaire completed by 413 subjects aged 25-74 yr in the Erie County Periodontal Disease (ECPD) Study. Specific questions on general and oral health conditions were completed by each subject during a first visit and at a follow-up examination 2 yr later, and the two compared. Results showed that the overall measure of agreement between the two visits is substantial (average kappa, kappa = 0.80). Variation by gender and age were minimal. Questions regarding allergy to medications, oral treatment, reason for tooth extraction, health symptoms and history of systemic diseases exhibited high levels of agreement (kappa ranged from 0.71-0.90). Information on vitamin and mineral intake yielded kappa = 0.63. Oral conditions scored the lowest but were still acceptable (kappa = 0.57). These findings indicate that there were no significant discrepancies in self-reported responses to the health questionnaire used in the ECPD Study. Although the information provided by the subject may not be as accurate as compared to laboratory testing, it is nevertheless a reliable source of information which can be utilized cost-effectively in research studies.

Estimating the burden of disease. Comparing administrative data and self-reports

OBJECTIVES

A cardiovascular health survey of a representative sample of the adult population of Manitoba, Canada was combined with the provincial health insurance claims database to determine the accuracy of survey questions in detecting cases of diabetes, hypertension, ischemic heart disease, stroke, and hypercholesterolemia.

METHODS

Of 2,792 subjects in the survey, 97.7% were linked successfully using a scrambled personal health insurance number. Hospital and physician claims were extracted for these individuals for the 3-year period before the survey.

RESULTS

The authors found no benefits to using restrictive criteria for entrance into the study (ie, requiring more than one diagnosis to define a case). Using additional years of data increased agreement between data sources. Kappa values indicated high levels of agreement between administrative data and self-reports for diabetes (0.72) and hypertension (0.59); kappa values were approximately 0.4 for the other conditions. Using administrative data as the "gold standard," specificity was generally very high, although cases with hypertension and hypercholesterolemia (diagnosed primarily by laboratory or physical measurement) were associated with a lower specificity than the other conditions. Sensitivity varied markedly and was lowest for "other heart disease" and "stroke". For diabetes and hypertension, inclusion criteria calling for more than one diagnosis reduced the accuracy of case identification, whereas increasing the number of years of data increased accuracy of identification. For diabetes and hypertension, self-reports were fairly accurate in detecting "true" past history of the illness based on physician diagnosis recorded on insurance claims.

CONCLUSIONS

This study demonstrates the feasibility of linking a large health survey with administrative data and the validity of self-reports in estimating the prevalence of chronic diseases, especially diabetes and hypertension. A linked data set offers unusual opportunities for epidemiologic and health services research in a defined population.

Self-rated health status as a health measure: the predictive value of self-reported health status on the use of physician services and on mortality in the working-age population

The validity of various self-reported health assessments in predicting physician contracts and all-cause mortality was investigated in a prospective study in Finland. The follow-up periods were one year for the use of physician services and ten years ten months for the mortality. The study cohort comprised 1340 men and 1500 women, 35-63 years of age at the beginning of the study. The initial health assessments were derived from postal questionnaires in 1980 (response rate 77.5%). The survey was repeated one year later to verify the stability of the respondents' perceived health status. The data on the physician contacts and mortality were registered independently. The stability of perceived health status was relatively good and the perceived health was inversely associated with the number of physician contacts per year. A consistent inverse association, standardized by age, sex and social status, was observed between perceived health status and perceived physical fitness and mortality, while the predictive value of self-reported chronic diseases was low. The results suggest that the subjective health assessments are valid health status indicator in middle-aged populations, and they can be used in cohort studies and population health monitoring.

Differences in the misreporting of chronic conditions, by level of education: the effect on inequalities in prevalence rates

OBJECTIVES

Many studies of socio-economic inequalities in the prevalence of chronic conditions rely on self-reports. For chronic nonspecific lung disease, heart disease, and diabetes mellitus, we studied the effects of misreporting on variations in prevalence rates by respondents' level of education.

METHODS

In 1991, a health interview survey was conducted in the southeastern Netherlands with 2867 respondents. Respondents' answers were compared with validated diagnostic questionnaires in the same survey and the diagnoses given by the respondents' general practitioners.

RESULTS

Misreporting of chronic lung disease, heart disease, and diabetes may be extensive. Depending on the condition and the reference data used, the confirmation fractions ranged between .61 and .96 and the detection fractions between .13 and .93. Misreporting varied by level of education, and although various patterns were observed, the dominant pattern was that of more underreporting among less educated persons. The effects on prevalence rates were to underestimate differences by level of education to a sometimes considerable degree.

CONCLUSIONS

Misreporting of chronic conditions differs by respondents' level of education. Health interview survey data underestimate socioeconomic inequalities in the prevalence of chronic conditions.

The elimination of selected chronic diseases in a population: the compression and expansion of morbidity

OBJECTIVES

This study evaluates the effect of eliminating a specific disease on the mortality, long-term disability, and overall health status of a population. Primarily, it examines whether elimination leads to a compression of morbidity.

METHODS

The Sullivan method was used to calculate disability-free life expectancy. Cause-deleted disability prevalence was estimated with a multiple logistic regression model that used data from the Dutch National Survey of General Practice. Cause-deleted probabilities of dying were derived with the cause-elimination life-table technique, assuming independence among competing causes of death.

RESULTS

Eliminating disabling nonfatal diseases such as arthritis/back complaints results in a decline in life expectancy with disability--that is, an absolute compression of morbidity. Eliminating highly fatal diseases such as cancer leads to an increase in the number of years and the proportion of life with disability--that is, a relative expansion of morbidity.

CONCLUSIONS

While eliminating fatal diseases leads to an increase in disability-free life expectancy, life expectancy with disability may increase as well. This represents an increasing burden to society. On the other hand, eliminating nonfatal disabling diseases leads to absolute compression of morbidity.

Rheumatoid factor, chronic arthritis and mortality

OBJECTIVE

To investigate chronic arthritis and rheumatoid factor (RF) for their prediction of premature total and cardiovascular mortality.

METHODS

In 1978-80, a representative population sample of 8000 Finns aged 30 or more was invited to participate in a comprehensive health examination; 90% complied. Arthritis was diagnosed on the basis of medical history, symptoms, and physical examination. Serum RF was determined by the sensitised sheep cell agglutination test.

RESULTS

By the end of 1992 1597 of the subjects had died from all causes, including 876 deaths from cardiovascular diseases. When adjusted for age, gender and smoking, the relative risk of persons with RF positive arthritis dying from any cause was 1.61 (95% confidence interval (CI) 1.03 to 2.51); RF negative non-erosive arthritis was not associated with mortality (relative risk 1.03; 95% CI 0.72 to 1.49). In the absence of arthritis, 'false positive' RF titres > or = 128 predicted cardiovascular deaths with a relative risk of 1.74 (95% CI 1.06 to 2.86).

CONCLUSION

Both RF positive arthritis and false positive RF reactions predict mortality, but through different disease patterns.

Recall bias in a case-control study of low birth weight

The role of report/recall bias in case-control studies of low birth weight (LBW) was investigated in women who gave birth at a tertiary hospital. Prenatal exposure information reported at the postpartum interview was compared with that documented during pregnancy in obstetric records. 169 cases of LBW and 198 controls were selected. The two sets of information on case mothers and control mothers were compared, using the medical record as a reference. Kappa values were estimated. No trend was observed to increase/decrease the sensitivity and specificity of recall. Agreement on alcohol use was very low (kappa = 0.11 for case mothers and 0.03 for control mothers): on obstetrical records, only 12 mothers of cases reported habitual alcohol intake at the first prenatal care visit, whereas in the interview 69 said yes to the same question; in control mothers, the figures were 4 and 89 respectively. Odds ratios (ORs) of exposure estimated from the two sets of data did not differ importantly in 8 variables. Interview data yielded ORs for hypertension (8.39 versus 4.63), anemia (0.44 versus 0.99) that were farther from the null, and ORs in the opposite direction for alcohol (0.83 versus 1.61) and any drug (0.64 versus 1.42). In conclusion, given that OR figures are similar for most variables and no trend is observed in sensitivity/specificity, mothers of normal births can be an adequate reference group, using personal interviews to obtain information on lifestyle, and medical records for conditions.

Methods of assessing prevalent cardiovascular disease in the Cardiovascular Health Study

The objective of this article is to describe the methods of assessing cardiovascular conditions among older adults recruited to the Cardiovascular Health Study (CHS), a cohort study of risk factors for coronary disease and stroke. Medicare eligibility lists from four US communities were used to obtain a representative sample of 5201 community-dwelling elderly, who answered standardized questionnaires and underwent an extensive clinic examination at baseline. For each cardiovascular condition, self-reports were confirmed by components of the baseline examination or, if necessary, by a validation protocol that included either the review of medical records or surveys of treating physicians. Potential underreporting of a condition was detected either by the review of medical records at baseline for other self-reported conditions or, during prospective follow-up, by the investigation of potential incident events. For myocardial infarction, 75.5% of the self-reports in men and 60.6% in women were confirmed. Self-reported congestive heart failure was confirmed in 73.3% of men and 76.6% of women; stroke, in 59.6% of men and 53.8% of women; and transient ischemic attack, in 41.5% of men and 37.0% of women. Underreporting was also common. During prospective follow-up of an average of about 3 years per person, approximately 50% of men and 38% of women were hospitalized or investigated for at least one potential incident event; for each cardiovascular condition, about 1 to 4% of those investigated during prospective follow-up were found to have had the cardiovascular condition prior to entry into the cohort.(ABSTRACT TRUNCATED AT 250 WORDS)