Journal of Diabetes and Its Comp

Medical records as sources of data on cardiovascular disease

events in persons with diabetesB

Barbara E.K. Kleina,T, Ronald Kleina, Patrick E. McBrideb,

Jennifer O. Reinkea, Michael D. Knudtsona

aDepartment of Ophthalmology and Visual Sciences, University of Wisconsin, Madison, WI 53726, USAbCardiovascular Medicine, University of Wisconsin, Madison, WI 53726, USA

Received 16 April 2004; received in revised form 16 June 2005; accepted 21 June 2005

Abstract

Purpose: The aim of this study is to evaluate medical records as a source of data on cardiovascular disease over a 20-year interval.

Methods: Participants in a population-based cohort of persons with Type 1 diabetes were asked whether they had been told by a doctor that

they had several specific cardiovascular events. In addition, they were asked when and where they were hospitalized for myocardial

infarction, stroke, surgical procedures, and for other conditions and procedures. The medical care institution was contacted to obtain copies of

the relevant hospitalization. Results: Overall, the confirmation of the self-reported events was 86.0% when medical records were obtained.

Percent confirmed variedwith the diagnosis. Reports of poor circulation in the lower extremities were confirmed in 42.6%, strokewas confirmed

in 70%, and coronary bypass surgerywas confirmed in 100% of cases. The success of obtainingmedical records was greater for those events that

were reported to have occurred more recently than those reported further in the past, especially when 10 or more years had elapsed.Conclusion:

Medical record confirmation of reported cardiovascular events in persons with Type 1 diabetes was high for some events when medical records

could be obtained but was lower for bpoor circulationQ to the legs and stroke possibly related to the lack of specificity of our questions, to

incorrect attribution of symptoms by the respondent, or to inaccurate recall of a physician’s examination.Medical record confirmation was better

for more recent than past events. Therefore, when hard copy documentation is needed, it should be sought within 10 years of the event.

D 2006 Elsevier Inc. All rights reserved.

Keywords: Cardiovascular disease; Cerebrovascular accident; Diabetes mellitus, Type 1; Medical records; Myocardial infarction

1. Introduction

Microvascular abnormalities (e.g., retinopathy and nephr-

opathy) are the hallmark of diabetes mellitus. It has been

suggested that large vessel disease in those with diabetes is

related to the antecedent microvascular abnormalities

(Morrish, Stevens, Fuller, Jarrett, & Keen, 1991; Yodaiken,

1976). We investigated this hypothesis in a population of

persons who had participated in the Wisconsin Epidemio-

1056-8727/06/$ – see front matter D 2006 Elsevier Inc. All rights reserved.

doi:10.1016/j.jdiacomp.2005.06.008

B This work was supported by grant HL59259.

T Corresponding author. Department of Ophthalmology and Visual

Sciences, University of Wisconsin, 610 North Walnut Street, Room 409

WARF, Madison, WI 53726, USA. Tel.: +1 608 263 0276; fax: +1 608

263 0279.

E-mail address: kleinb@epi.ophth.wisc.edu (B.E.K. Klein).

logic Study of Diabetic Retinopathy (Klein et al., 2004).

Medical records and self-reports are frequently utilized to

ascertain outcomes in medical research. The accuracy of self-

reports is critical to the methodology of long-term studies of

health outcomes. It is becoming increasingly difficult to

access relevant medical records. We describe here our

experience with obtaining medical record data to document

macrovascular events in a population-based study of persons

with long duration diabetes.

2. Methods

Institutional review board approval was obtained for all

participants at each phase of the Wisconsin Epidemiologic

Study of Diabetic Retinopathy, and individual informed

lications 20 (2006) 224–227

Table 1

Cardiovascular events questionnaire

Have you ever been told by a physician that you had angina—that is, pressure in the chest on exertion due to heart disease? Yes, No, Don’t Know

Have you ever been told by a physician that you have had a heart attack? Yes, No, Don’t Know

Have you had a coronary bypass? Yes, No, Don’t Know

Have you had a coronary angioplasty or stent? Yes, No, Don’t Know

Have you ever been told by a physician that you have had a stroke (apoplexy, cerebral vascular accident)? Yes, No, Don’t Know

Have you had surgery to the brain or neck to prevent or correct a stroke? Yes, No, Don’t Know

Have you ever had a doctor or other medical person tell you that you have poor circulation to legs or feet? Yes, No, Don’t Know

Have you had a toe(s) or foot surgically amputated due to infection or poor circulation? Yes, No, Don’t Know

Have you had a leg surgically amputated due to infection or poor circulation? Yes, No, Don’t Know

Have you ever had surgery to your hands, legs, or feet for circulation problems other than for varicose veins

(or an amputation or accident or carpal tunnel)?

Yes, No, Don’t Know

How many times have you been hospitalized overnight or longer since we last spoke with you on (Date of last interview)? Count or Don’t Know

Are any of these hospitalizations in addition to what you have already reported? Yes, No, Don’t Know

Table 2

Medical records and confirmation of cardiovascular events

Event

Number

of records

requested

Number of

received (% of

those requested)

Number of

confirmation (%

of those received)

All events 837 615 (73.5) 529 (86.0)

Myocardial infarction 71 57 (80.3) 50 (87.7)

Coronary

angioplasty/stent

46 37 (80.4) 34 (91.9)

Poor circulation to legs 164 108 (65.9) 46 (42.6)

Amputation, toe 38 28 (73.7) 27 (96.4)

Amputation, leg 23 20 (87.0) 19 (95.0)

Angina 78 58 (74.4) 46 (79.3)

Stroke 23 20 (87.0) 14 (70.0)

Operation to

prevent stroke

3 3 (100) 2 (66.7)

Circulatory

operation on legs

28 23 (82.1) 16 (69.6)

Coronary artery bypass 55 49 (89.1) 49 (100)

Miscellaneous

hospitalizations

308 227 (73.7) 226 (99.6)

B.E.K. Klein et al. / Journal of Diabetes and Its Complications 20 (2006) 224–227 225

consent was obtained for each visit. In 1979–1980, all

persons with diabetes diagnosed before they were 30 years

of age, who were taking insulin, and who were receiving

care in an 11-county area of southern Wisconsin were

identified (Klein et al., 1984). There were 1210 such

persons; 996 of them participated at baseline in a study of

the prevalence of diabetic retinopathy. The initial study

evaluation included a medical history, measurements of

blood pressures, glycosylated hemoglobin, height, weight,

urinalysis, dilation of the pupil of the eye, and the taking of

seven standard photographic fields of the ocular fundus of

each eye. Participants responded to a medical history

questionnaire at each examination. Pertinent to the current

investigation were responses to interviews conducted

between 1996 and 2001, when study participants were

asked whether a physician had told them that they had had

angina, a heart attack, stroke, poor circulation in their legs,

coronary angioplasty, coronary bypass, surgery in their head

or neck to prevent a stroke, or surgery for circulation in their

legs. For each positive response, we recorded the date and

place of the relevant hospitalization for the condition if they

were hospitalized for it. In addition, we asked for the date

and name of the hospital for any additional hospitalizations

(Table 1). We requested consent from each person reporting

an event to obtain the medical record for each relevant event

and for miscellaneous hospitalizations. All records received

were reviewed by one of the authors (JOR) for correctness

of name of participant and hospital, date of reported event,

and so on. When discrepancies were detected, hospitals

were recontacted up to three times in an attempt to obtain

the appropriate record. These medical records were

reviewed using standard protocols adapted from those

developed for the Women’s Health Initiative (Section 2;

Clinical Centers for the Clinical Trial and Observational

Study of the Women’s Health Initiative East/West. WHI

Protocol Manual, 1997; Writing Group for the Women’s

Health Initiative Investigators, 2002) by a single investi-

gator (PEM). The review protocol described electrocardio-

gram characteristics, cardiac enzyme levels, specific

characteristics of stroke, recorded descriptions of chest pain

for angina and leg pain for the diagnosis of poor circulation

in the legs, and for the procedures listed above. The

reviewer was masked as to the self-report. Miscellaneous

hospitalizations were defined as overnight hospitalizations

for reasons other than cardiovascular events.

During the 20-year follow-up examination, electrocardio-

grams were obtained according to the EPICARE protocol

(EPICARE Center of the Department of Public Health and

Sciences of Wake Forest University School of Medicine).

They were graded by the Nova code (Rautaharju et al., 1990).

3. Results

We obtained interview data from 704 of the original

cohort; 273 had died, and 19 could not be located. The mean

age of our study participants at the time of requesting the

records was 40.8 (F11.1) years, with a range of 18–80 years

of age. We confined these analyses to the first reported event

of each type in the 704 persons. Records were requested and

reviewed for 837 events, and other hospitalizations and were

reviewed between 1997 and 2001. Table 2 describes the

B.E.K. Klein et al. / Journal of Diabetes and Its Complications 20 (2006) 224–227226

results of these efforts for all cardiovascular events and for

each individual endpoint. Six hundred and fifteen of the 837

requested records were received (73.5%). For all endpoints,

529 (86.0%) were confirmed by record review. The percent

of records received of those requested differed by diagnosis

from 65.9% of those for poor circulation in the legs to 89.1%

of those for coronary bypass surgery.

The most commonly reported endpoints were poor

circulation to the legs and angina. For the former, 46/108

(42.6%) was confirmed and for the latter, 46/58 (79.3%) was

confirmed by record review. Percent confirmed differed by

diagnosis for other endpoints as well. For coronary events

aside from angina, the confirmation rate was, in general, high

ranging from 87.7% for myocardial infarction to 100% for

coronary artery bypass surgery. Reports of miscellaneous

hospitalizations were highly accurate because there were no

specific procedures or diagnostic tests necessary to confirm

this, only a discharge summary or progress note that

demonstrated the person was admitted and kept overnight.

The participant was not asked to recall any details other than

the year (give or take one year) and institution.

Table 3 describes the percent confirmed by the time

interval between the interview date (1996–2001) and the

date of the event (1978–2001). The proportion of records

received decreased slightly with duration up to 9 years; the

proportion was much decreased after 10 or more years. The

proportion of confirmations of self-reported event when

records were received was greater for events more recent

than 10 years compared with those occurring 10 or more

years before. There were too few of most events to estimate

this effect for each endpoint. There was no evidence to

suggest that any particular event was more likely to have

occurred at a different duration from the time of record

request than any other event.

Reasons for lack of confirmation included participants’

refusal to consent or to give complete information relating to

the particular event (4.1% of all reported events), medical

facility did not send records in response to our inquiries

(these were repeated up to three times; 20.7%), or the

records sent were incomplete (12.1%).

There were nine cases of definite or probable myocardial

infarction (Nova codes 5.1 or 5.2) among the participants

who had electrocardiograms (n=557) at the 20-year exam.

Seven of these occurred in those 71 persons who had

reported myocardial infarction.

Table 3

Time between event and when self-reported by confirmation status

Duration

(years)

Number of

records requested

Number of records

received (% of

those requested)

Number confirmed

(% of those received)

0–1 309 243 (78.6) 203 (83.5)

2–4 343 260 (75.8) 226 (86.9)

5–9 112 83 (74.1) 71 (85.5)

10+ 72 44 (61.1) 29 (65.9)

4. Discussion

Self-reported history of specific diseases has been found

to be reliable and useful in other studies of several different

morbidities (Ivers, Cumming, Mitchell, & Peduto, 2002;

Sangha, Stucki, Liang, Fossel, & Katz, 2003; Tormo,

Navarro, Chirlaque, & Barber, 2000). However, the reli-

ability varies with the specific disease (Bergmann et al.,

1998; Desai, Bruce, Desai, & Druss, 2001; Klungel, de Boer,

Paes, Seidell, & Bakker, 1999). We have found that

participant-reported history of a doctor-diagnosed coronary

heart disease, including a variety of categories in persons

with Type 1 diabetes, was high. In the Cardiovascular Health

Study (CHS), at the baseline examination, 16.6% of men and

8.3% of women reported a history of myocardial infarction

(Psaty et al., 1995). This was confirmed for 76.6% for men

and 65.1% for women. For persons not reporting a

myocardial infarction, 4.4% of men and 3.2% of women

had electrocardiographic evidence of a previous myocardial

infarction. We note that the age range in the CHS was 65 and

older (Fried et al., 1991). The younger age of our study

participants may have been associated with the somewhat

better accuracy of our self-reports. We did not attempt to

obtain medical records for study participants until the

20-year follow-up of our cohort, and even then, we only

had resources to follow-up for reported events (including

bmiscellaneousQ hospitalizations). Nonetheless, we were

able to confirm 50 of the 71 self-reported cases of

myocardial infarction. The proportion of cases confirmed

varied with the specific cardiovascular diagnosis. The

percent confirmed was poorest for poor circulation in the

legs (42.6%). This likely reflects the lack of etiologic

specificity for the symptoms of numbness, tingling, and loss

of sensation in the lower extremity which could be attributed

to neuropathy or impaired circulation. For this diagnosis,

there were often cases (53%) when peripheral angiography

or ultrasound was performed and no pathology of vessels or

measurable peripheral pulses were noted. It is possible that

self-report of poor circulation reflects errors in patient’s

attribution of the cause of symptoms or differences in

definition of bpoor circulation.Q Our questionnaire did not

include queries as to specific symptoms or signs of bpoorcirculation.Q Even if we had included questions aimed at

distinguishing between bpoor circulationQ and neuropathy, it

is uncertain that the study participants or we would be likely

to classify the cause of such symptoms correctly. It is also

unclear as to whether the physician’s explanation to a

participant or that individual’s remembrance of such a

discussion would be accurate.

We were hampered in our ability to retrieve medical

records because in some cases records were destroyed by

hospitals and in some cases hospitals no longer existed. In

some cases, the relatively long interval between the event

and our attempts to retrieve the records likely accounted for

the 222 (27%) cases when hospitals did not return requested

records. In addition, in a few cases, participants refused to

B.E.K. Klein et al. / Journal of Diabetes and Its Complications 20 (2006) 224–227 227

give consent for us to obtain the records or hospitals refused

to release them. Recent changes in privacy regulations and

the climate of trust may make it more difficult to obtain

such records in the future causing researchers to depend

more on self-report. While self-reporting of a doctor’s

diagnosis may lead to misclassification, so might depending

on retrieval of appropriate medical records. We note that the

recency of event was associated with a higher percentage of

obtainable records, and so if medical record confirmation is

critical, this should be attempted as soon as possible after

the event is reported.

We obtained electrocardiograms at the 20-year follow-up

examination. The number of persons with past myocardial

infarction according to Nova code criteria was only nine.

This suggests that this is not a good measure of the burden

of myocardial infarction in those with Type 1 diabetes.

In summary, this study demonstrates that patient self-

reports of cardiovascular disease events are correlated with

actual documented cardiovascular heart disease events, but

the strength of the correlation varies with the specific event.

However, if the event is more than 10 years ago, the

reliability of the patient self-report and availability of related

medical records drops significantly. Therefore, participants

in prospective studies should be urged to update their health

histories and report any new events. Retrospective reviews

are most accurate if the event has occurred within the prior

decade. Substantial effort is required on the part of research

staff to obtain medical documentation of clinical events.

Acknowledgments

The authors thank Lisa Grady for technical support of

this manuscript.

References

Bergmann, M. M., Calle, E. E., Mervis, C. A., Miracle-McMahill, H. L.,

Thun, M. J., & Heath, C. W. (1998). Validity of self-reported cancers in

a prospective cohort study in comparison with data from state cancer

registries. American Journal of Epidemiology, 147, 556–562.

Clinical Centers for the Clinical Trial and Observational Study of the

Women’s Health Initiative East/West. WHI Protocol Manual, Vol. 8—

Outcomes, pp. 1–7, Version 8/15/1997.

Desai, M. M., Bruce, M. L., Desai, R. A., & Druss, B. G. (2001). Validity

of self-reported cancer history: A comparison of health interview data

and cancer registry records. American Journal of Epidemiology, 153,

299–306.

Fried, L. P., Borhani, N. O., Enright, P., Furberg, C. D., Gardin, J. M.,

Kronmal, R. A., Kuller, L. H., Manolio, T. A., Mittelmark, M. B.,

Newman, A., et al. (1991). The Cardiovascular Health Study: Design

and rationale. Annals of Epidemiology, 1, 263–276.

Ivers, I. Q., Cumming, R. G., Mitchell, P., & Peduto, A. J. (2002). The

accuracy of self-reported fractures in older people. Journal of Clinical

Epidemiology, 55, 452–457.

Klein, B. E., Klein, R., McBride, P. E., Cruickshanks, K. J., Palta, M.,

Knudtson, M. D., et al. (2004). Cardiovascular disease, mortality, and

retinal microvascular characteristics in type 1 diabetes: Wisconsin

Epidemiologic Study of Diabetic Retinopathy. Archives of Internal

Medicine, 164, 1917–1924.

Klein, R., Klein, B. E., Moss, S. E., DeMets, D. L., Kaufman, I., & Voss, P.

S. (1984). Prevalence of diabetes mellitus in southern Wisconsin.

American Journal of Epidemiology, 119, 54–61.

Klungel, O. H., de Boer, A., Paes, A. H., Seidell, J. C., & Bakker, A.

(1999). Cardiovascular diseases and risk factors in a population-

based study in The Netherlands: Agreement between questionnaire

information and medical records. Netherlands Journal of Medicine,

55, 177–183.

Morrish, N. J., Stevens, L. K., Fuller, J. H., Jarrett, R. J., & Keen, H.

(1991). Risk factors for macrovascular disease in diabetes mellitus: The

London follow-up to the WHOMultinational Study of Vascular Disease

in Diabetics. Diabetologia, 34, 590–594.

Psaty, B. M., Kuller, L. H., Bild, D., Burke, G. L., Kittner, S. J., Mittelmark,

M., et al. (1995). Methods for assessing prevalent cardiovascular

disease in the Cardiovascular Health Study. Annals of Epidemiology,

5, 270–277.

Rautaharju, P. M., MacInnis, P. J., Warren, J. W., Wolf, H. K., Rykers, P.

M., & Calhoun, H. P. (1990). Methodology of ECG interpretation in the

Dalhousie program: NOVA CODE ECG classification procedures for

clinical trials and population health surveys. Methods of Information in

Medicine, 29, 262–274.

Sangha, O., Stucki, G., Liang, M. H., Fossel, A. H., & Katz, J. N. (2003).

The Self-Administered Comorbidity Questionnaire: A new method to

assess comorbidity for clinical and health services research. Arthritis

and Rheumatism, 49, 156–163.

Tormo, M. J., Navarro, C., Chirlaque, M. D., & Barber, X. (2000).

Validation of self diagnosis of high blood pressure in a sample of

the Spanish EPIC cohort: Overall agreement and predictive values.

EPIC Group of Spain. Journal of Epidemiology and Community

Health, 54, 221–226.

Writing Group for the Women’s Health Initiative Investigators. (2002).

Risks and benefits of estrogen plus progestin in healthy postmenopausal

women. Principal results from the Women’s Health Initiative Random-

ized Controlled Trial. Journal of the American Medical Association,

288, 321–333.

Yodaiken, R. E. (1976). The relationship between diabetic capilloropathy

and myocardial infarction: A hypothesis. Diabetes, 25, 928–930.